Skip to content
The AI agent playbook

Learn how AI agents fit into your team’s work, step by step

Eighteen everyday workflows from six teams, drawn step by step. Each one shows which steps plain automation handles, which ones an AI agent can take on, and the calls that stay with your people.

Workflows from different teams drawn as three lanes, workflow automation, an AI agent and a person, with a job travelling through each one from what starts it to the finished result.
How to read a workflow

Every workflow has three kinds of work

Most steps are plain automation. A few need someone to read something and decide, which is where an AI agent comes in. And some calls should always stay with a person.

  • Workflow automation

    Moves the work.

    Copies the record into the next system, sends the reminder, books the meeting. Your team has been able to automate this part for years.

  • AI agent

    Reads and decides.

    Checks the offer against the pay band, matches the invoice to the order, drafts the reply. The steps that used to wait for someone to look.

  • A person

    Makes the call.

    Approves the salary outside the band, signs off the refund over the limit. The agent hands over everything needed to decide quickly.

The workflows

Pick a workflow. See who does what.

Every map runs from what starts the work to the finished result. Select a step to see what happens there, which systems it touches, and why it sits in that lane.

Finance AI for Finance →

Invoice to payment

Paying vendor bills on time, and only once

Starts with

Invoice arrives

A vendor emails an invoice to the AP inbox as a PDF, a scan or a phone photo.

AI agent: Takes the vendor, invoice number, each line, the tax and the total off the page, whether it came as a tidy PDF or a phone photo.
Workflow automation: Looks up the vendor and invoice number in accounting. A second copy of INV-4471 stops there, and anything new is logged as a bill.
What if: 83 hours a month back, worth $56,800 a year

What if you had it

83 hours a month back$56,800 a year

About 5 minutes of keying on each of 1,000 invoices, plus $12,000 in duplicate payments stopped.

Customize to your business
Starts when
A vendor emails an invoice to the AP inbox as a PDF, a scan or a phone photo.
Result
Each invoice logged once and never paid twice.

Why each part sits where it does

  • AI agent. Each vendor puts the total somewhere different, and some never label it at all.
  • Workflow automation. A duplicate has the same vendor and the same invoice number. An exact match is enough to catch it.

Reads

  • AP inbox
  • Vendor records

Writes

  • Accounting

The agent

Invoice matchingMatches invoices to POs and receipts
Step 1 of 5
AI agent: Compares each line with the purchase order and what the warehouse signed for, on price and quantity, even when the vendor names a part its own way.
What if: 67 hours a month back, worth $36,200 a year

What if you had it

67 hours a month back$36,200 a year

About 4 minutes of line checks on each of 1,000 invoices.

Customize to your business
Starts when
Each invoice logged once and never paid twice.
Result
Every line matched, or the gap named in dollars.

Why each part sits where it does

  • AI agent. The vendor bills “HD bracket, zinc” and your PO says “Mounting bracket 40mm”. A rule sees two different parts.

Reads

  • Purchase orders
  • Receiving log

Writes

  • Accounting

The agent

Invoice matchingMatches invoices to POs and receipts
Step 2 of 5
Workflow automation: Clean matches post as draft bills. Anything past the $50 tolerance goes to the controller with the invoice, PO and receipt on one screen.
If over the $50 tolerancePerson: Approves the extra $212 on PO 2208, or turns it down and asks the vendor for a corrected invoice.Declined: back to the vendor for a credit
What if: 27 hours a month back, worth $14,600 a year

What if you had it

27 hours a month back$14,600 a year

About 20 minutes of gathering paperwork on each of 80 exceptions.

Customize to your business
Starts when
Every line matched, or the gap named in dollars.
Result
Clean bills wait as drafts, and the one exception has a name on it.

Why each part sits where it does

  • Workflow automation. The tolerance is a set number. Which bills need a person is known before anyone opens one.
  • Person. Paying above the PO is new spending, and someone with the authority to approve it has to own it.

Reads

  • Purchase orders

Writes

  • Accounting
Step 3 of 5
Workflow automation: Pulls every approved bill due before the next run, and moves up any bill where paying ten days early earns a 2% discount.
Person: Checks the total and any vendor whose bank details changed since its last payment, then sends the run to the bank.Declined: held, with the bills flagged
What if: 10 hours a month back, worth $23,400 a year

What if you had it

10 hours a month back$23,400 a year

About 10 hours a month building runs and spotting discount windows, plus $18,000 in early-payment discounts taken.

Customize to your business
Starts when
Clean bills wait as drafts, and the one exception has a name on it.
Result
Bills paid on their due date, or early when the discount is worth it.

Why each part sits where it does

  • Workflow automation. Due dates and discount windows are just dates. Sorting by them takes no judgment.
  • Person. Money leaving the account is where your controls already put a person, and a changed bank detail deserves a second look.

Reads

  • Accounting
  • Vendor records

Writes

  • Bank
Step 4 of 5
Workflow automation: Once the bank confirms, marks the bill paid, emails the remittance advice and closes the PO.Remittance emailBill marked paidPO closed
AI agent: Replies to vendor payment questions with the bill’s status, pay date and reference. A disputed charge goes to AP instead.
What if: 16 hours a month back, worth $8,600 a year

What if you had it

16 hours a month back$8,600 a year

About 8 minutes on each of 120 vendor payment questions.

Customize to your business
Starts when
Bills paid on their due date, or early when the discount is worth it.
Result
Vendors hear back the same day, and AP never has to open the ledger.

Why each part sits where it does

  • Workflow automation. Every paid bill needs the same three updates, whoever the vendor is.
  • AI agent. “Any news on the March invoice?” could mean four different bills, and the vendor quotes its own number, not yours.

Reads

  • Accounting
  • AP inbox

Writes

  • Email
  • Purchase orders

The agent

Vendor inquiriesAnswers vendor payment questions from the ledger
Step 5 of 5

Ends with

Paid on time

The bill is paid by its due date, the PO is closed and the vendor already has the remittance.

Total203h/mo$140k/yr

What changes

AP stops retyping invoices. The controller sees only the bills that don’t match the PO, with the paperwork beside each one.

  • 4steps automated
  • 3steps for an AI agent
  • 2calls kept with a person

What if you had it

203 hours a month back$139,600 a year in time and savings

For a 400-person company paying about 1,000 vendor invoices a month, at $45 an hour.

Customize to your business

Month-end close

A close where every number comes with its reason

Starts with

Month ends

The last business day of the month passes and the close calendar starts.

Workflow automation: Copies the close checklist into the new month, dates each task by working day and assigns it to its usual owner.
AI agent: Reads last month’s notes, new contracts and open items for anything the template misses, like a new office lease that needs its first entry.
What if: 8 hours a month back, worth $6,200 a year

What if you had it

8 hours a month back$6,200 a year

About 8 hours rebuilding the list and hunting for new items.

Customize to your business
Starts when
The last business day of the month passes and the close calendar starts.
Result
This month’s list, with an owner and a date on each task.

Why each part sits where it does

  • Workflow automation. Most of the close is the same list in the same order, month after month.
  • AI agent. A new lease or a one-off bonus turns up in a contract or an email long before anyone adds it to the template.

Reads

  • Accounting
  • Shared drive

Writes

  • Close checklist

The agent

Close checklistTracks every close task and its evidence
Step 1 of 5
Workflow automation: Pulls the month’s balances from the sub-ledgers, the general ledger and the bank, and compares them account by account.
AI agent: Where balances disagree, works through the transactions on both sides to find why, such as $540 of card fees the bank took and nobody booked.
What if: 60 hours a month back, worth $66,800 a year

What if you had it

60 hours a month back$66,800 a year

About 90 minutes on each of 40 account reconciliations, plus $20,000 in close overtime and temp help.

Customize to your business
Starts when
This month’s list, with an owner and a date on each task.
Result
Accounts tied, and each gap traced to the entry behind it.

Why each part sits where it does

  • Workflow automation. Two balances either agree or they don’t, and subtraction says which.
  • AI agent. The gap is one number. The reason is buried in hundreds of transactions on both sides.

Reads

  • Sub-ledgers
  • Accounting
  • Bank

Writes

  • Close checklist

The agent

Three-way reconciliationTies sub-ledger, ledger and bank daily
Step 2 of 5
AI agent: Drafts the accruals and an entry for each gap, treating each item the way it was treated last period, with the support attached.
Person: Reads each entry with its support, like the $540 booked to bank fees, and approves it or sends it back with a question.Declined: back to the preparer with a question
Workflow automation: Posts approved entries to the ledger, with the support and the approver’s name on each one.
What if: 30 hours a month back, worth $23,400 a year

What if you had it

30 hours a month back$23,400 a year

About 45 minutes drafting each of 40 adjusting entries.

Customize to your business
Starts when
Accounts tied, and each gap traced to the entry behind it.
Result
Adjustments posted, each with its reason and approver on the entry.

Why each part sits where it does

  • AI agent. Whether to accrue something is a judgment call. Making it the same way every month is what keeps the books comparable.
  • Person. Your controls require a named approver on every journal entry, and this is where they sit.
  • Workflow automation. Once an entry is approved, posting it is data entry.

Reads

  • Accounting
  • Shared drive

Writes

  • Accounting

The agent

Accrual and treatment checkApplies the same test to every entry
Step 3 of 5
Workflow automation: Once the entries are in, runs the month’s actuals against budget for each department.
If over 10% or $5,000AI agent: Writes a sentence on each big variance from the entries behind it: a software renewal paid early, two hires who started a month late.
What if: 24 hours a month back, worth $18,700 a year

What if you had it

24 hours a month back$18,700 a year

About 3 hours of commentary for each of 8 departments.

Customize to your business
Starts when
Adjustments posted, each with its reason and approver on the entry.
Result
Each big variance has a sentence saying why it moved.

Why each part sits where it does

  • Workflow automation. Actuals against budget is a report you already have. It only needs running on the right day.
  • AI agent. The total shows that a line moved. The invoices and entries underneath say why.

Reads

  • Accounting
  • Budget

Writes

  • Shared drive

The agent

Budget vs actualExplains each month’s variances to budget
Step 4 of 5
AI agent: Opens the support on every task marked done and flags any that don’t back it up, like a reconciliation still dated last month.
Person: Reviews the variance notes and the finished checklist, asks about anything that looks off, then signs off the month.Declined: reopened, with the tasks named
Workflow automation: Locks the period in accounting, sends the close pack to leadership and files the support for the auditors.Period lockedPack to leadershipSupport filed
What if: 12 hours a month back, worth $9,400 a year

What if you had it

12 hours a month back$9,400 a year

About 6 minutes checking the support on each of 120 close tasks.

Customize to your business
Starts when
Each big variance has a sentence saying why it moved.
Result
The month closed, with its support in one place for the auditors.

Why each part sits where it does

  • AI agent. A task ticked with last month’s file behind it looks the same as one that was done. Only opening the file shows it.
  • Person. The numbers go out under the CFO’s name.
  • Workflow automation. Locking the period and sending the pack are a few fixed clicks at the end of every close.

Reads

  • Close checklist
  • Shared drive

Writes

  • Accounting
  • Email
  • Shared drive

The agent

Close checklistTracks every close task and its evidence
Step 5 of 5

Ends with

Books signed off

Every account tied, every task backed by its file, and each big variance explained before leadership asks.

Total134h/mo$124k/yr

What changes

By the first morning of close the balances are tied and each gap is traced, leaving the team the questions that need a controller.

  • 5steps automated
  • 5steps for an AI agent
  • 2calls kept with a person

What if you had it

134 hours a month back$124,500 a year in time and savings

For a finance team of eight closing the books each month for a company of about 400 people, at $65 an hour.

Customize to your business

Expense reports

Employee expenses checked, coded and paid back

Starts with

Report submitted

An employee submits the month’s expenses with photos of the receipts.

AI agent: Takes the merchant, date, amount and tax off every photo, including faded thermal paper and a hotel bill in euros.
Workflow automation: Finds the card charge with the same amount and date for each receipt, and converts foreign spend at that day’s published rate.
What if: 30 hours a month back, worth $18,000 a year

What if you had it

30 hours a month back$18,000 a year

About 6 minutes matching receipts to card charges on each of 300 reports.

Customize to your business
Starts when
An employee submits the month’s expenses with photos of the receipts.
Result
Every line backed by a receipt and a card charge.

Why each part sits where it does

  • AI agent. A photo of a taxi receipt is a picture, with no fields in it until someone reads it.
  • Workflow automation. Matching on amount and date is exact, and the exchange rate for any day is published.

Reads

  • Expense tool
  • Company cards

Writes

  • Expense tool

The agent

Expense report reviewChecks receipts against your policy
Step 1 of 5
AI agent: Tests every line against your written policy: the $400 meals limit, the hotel cap, alcohol, and whether a “client dinner” names the client.
What if: 25 hours a month back, worth $24,000 a year

What if you had it

25 hours a month back$24,000 a year

About 5 minutes of policy checks on each of 300 reports, plus $9,000 in out-of-policy spend caught.

Customize to your business
Starts when
Every line backed by a receipt and a card charge.
Result
Lines inside the policy cleared, the rest flagged with the rule they break.

Why each part sits where it does

  • AI agent. Your policy is written in sentences. Whether a dinner counts as client entertainment depends on the receipt and the note read together.

Reads

  • Expense policy
  • Expense tool

Writes

  • Expense tool

The agent

Expense report reviewChecks receipts against your policy
Step 2 of 5
Workflow automation: Sends the report to the employee’s manager, and adds the finance manager for any line that goes over a limit.
If over a limitPerson: Approves the $640 client dinner that went over the $400 meals limit, or declines it with a note the employee can see.Declined: back to the employee with the reason
What if: 20 hours a month back, worth $12,000 a year

What if you had it

20 hours a month back$12,000 a year

About 4 minutes of prep saved for approvers on each of 300 reports.

Customize to your business
Starts when
Lines inside the policy cleared, the rest flagged with the rule they break.
Result
Approved, with every exception signed by name.

Why each part sits where it does

  • Workflow automation. Who approves is set by the org chart and the size of the line.
  • Person. An exception is a decision about what the company will pay for, and your policy names who makes it.

Reads

  • Org chart
  • Expense policy

Writes

  • Expense tool
Step 3 of 5
AI agent: Picks the account, department and project for each line, so a lunch with a client lands on that client’s project and not on office costs.
Workflow automation: Posts the approved, coded lines to accounting as a single entry, with every receipt attached.
What if: 25 hours a month back, worth $15,000 a year

What if you had it

25 hours a month back$15,000 a year

About 5 minutes of account coding on each of 300 reports.

Customize to your business
Starts when
Approved, with every exception signed by name.
Result
Booked to the right accounts, with the receipts on the entry.

Why each part sits where it does

  • AI agent. Coffee for a client workshop and coffee for the office go to different accounts. Only the receipt and the note say which.
  • Workflow automation. Once every line has a code, the entry is the lines added up.

Reads

  • Expense tool
  • Accounting

Writes

  • Accounting

The agent

Expense report reviewChecks receipts against your policy
Step 4 of 5
Workflow automation: Puts out-of-pocket spend on the next paycheck, closes the matched card charges and tells the employee.Next payrollCard lines closedEmployee notice
What if: 8 hours a month back, worth $4,800 a year

What if you had it

8 hours a month back$4,800 a year

About 8 hours a month on reimbursements and card cleanup.

Customize to your business
Starts when
Booked to the right accounts, with the receipts on the entry.
Result
Money back on payday, with no one chasing it.

Why each part sits where it does

  • Workflow automation. An approved amount is one more line on the next payroll, like any other.

Reads

  • Expense tool

Writes

  • Payroll
  • Company cards
  • Email
Step 5 of 5

Ends with

Paid back and booked

Money back on the next payroll, every line coded to the right account and every card charge matched.

Total108h/mo$74k/yr

What changes

Out-of-policy lines reach finance with the rule they break. Everyone else is paid back on the next payroll without asking.

  • 4steps automated
  • 3steps for an AI agent
  • 1call kept with a person

What if you had it

108 hours a month back$73,800 a year in time and savings

For a 400-person company where about 300 people file an expense report each month, at $50 an hour.

Customize to your business

Sales & Marketing AI for Sales & Marketing →

Lead to the right rep

A brief and a reply for every demo request

Starts with

Demo request comes in

Someone asks for a demo through the form on your website.

Workflow automation: Creates the lead in the CRM, tagged with the page they filled the form in on and the campaign that brought them there.
AI agent: “Harbor and Pine” and “Harbor & Pine Ltd” become one company, and if that company has an open deal, the lead joins it.
What if: 30 hours a month back, worth $23,400 a year

What if you had it

30 hours a month back$23,400 a year

About 6 minutes logging and deduping each of 300 leads.

Customize to your business
Starts when
Someone asks for a demo through the form on your website.
Result
One clean lead, on the right company.

Why each part sits where it does

  • Workflow automation. Each form field has one place to go in the CRM, every time.
  • AI agent. A matching rule needs identical spellings, and nobody types a company name the same way twice.

Reads

  • Web form
  • CRM

Writes

  • CRM

The agent

Inbound lead briefResearches and routes every new lead
Step 1 of 5
AI agent: Reads their website and what they typed on the form, then writes a brief a rep can scan in a minute: headcount, systems, the likely need.
What if: 75 hours a month back, worth $58,500 a year

What if you had it

75 hours a month back$58,500 a year

About 15 minutes researching each of 300 companies.

Customize to your business
Starts when
One clean lead, on the right company.
Result
The brief is on the record before the rep opens it.

Why each part sits where it does

  • AI agent. Headcount comes from a lookup. What they probably need is in their own words on the form, and on their site.

Reads

  • CRM
  • Company website

Writes

  • CRM

The agent

Inbound lead briefResearches and routes every new lead
Step 2 of 5
AI agent: Weighs the lead against the customer profile you wrote down: company size, industry, the need in the brief and which pages they visited.
If over 200 peoplePerson: Reads the brief and the score, and decides whether a large account goes to the enterprise team.Declined: stays with the territory rep
Workflow automation: Sets the owner from your territory rules and the score, and sends that rep the brief.Enterprise teamTerritory repNurture
What if: 25 hours a month back, worth $19,500 a year

What if you had it

25 hours a month back$19,500 a year

About 5 minutes scoring and routing each of 300 leads.

Customize to your business
Starts when
The brief is on the record before the rep opens it.
Result
The lead has an owner within the hour.

Why each part sits where it does

  • AI agent. A points table adds up size, need and behavior one at a time. Judging fit means weighing them against each other.
  • Person. One big account can change a team’s quarter. Your routing rule puts those in front of the sales lead first.
  • Workflow automation. Once the score is set, the owner is one lookup in the territory rules.

Reads

  • CRM
  • Territory rules

Writes

  • CRM
  • Email

The agent

Lead scoring and routingScores fit and sends leads to the right rep
Step 3 of 5
AI agent: Answers the question they asked on the form, in the rep’s own tone, and offers two open times from the rep’s calendar next week.
Person: Adds what they know that the CRM doesn’t, like a mutual contact, and sends it from their own inbox.
What if: 40 hours a month back, worth $31,200 a year

What if you had it

40 hours a month back$31,200 a year

About 8 minutes writing each of 300 first replies.

Customize to your business
Starts when
The lead has an owner within the hour.
Result
A first reply from a named rep, not a no-reply address.

Why each part sits where it does

  • AI agent. A template can fill in a first name. It can’t answer the question this person actually typed.
  • Person. The rep who will run the deal should be the one who says hello.

Reads

  • CRM
  • Calendars

Writes

  • Email
  • CRM

The agent

Inbound lead briefResearches and routes every new lead
Step 4 of 5
AI agent: Looks at what each lead has opened, clicked and read, then chooses their next email and adjusts it to where they are.
Workflow automation: Sends on schedule, logs opens and clicks in the CRM, and pings the owner when a lead revisits pricing.SendCRM logRep alert
What if: 13 hours a month back, worth $10,100 a year

What if you had it

13 hours a month back$10,100 a year

About 4 minutes choosing the next email for each of 200 leads.

Customize to your business
Starts when
A first reply from a named rep, not a no-reply address.
Result
When a quiet lead warms up, the owner hears about it.

Why each part sits where it does

  • AI agent. Two people from the same webinar can be a month apart in what they need to hear next.
  • Workflow automation. Each open or click is an event with a fixed response: send, log or alert.

Reads

  • Marketing email
  • CRM

Writes

  • Marketing email
  • CRM
  • Chat

The agent

Lead nurture sequencesTailors follow-ups to what each lead does
Step 5 of 5

Ends with

Owned the same day

The right rep owns the lead and has replied, and leads who aren’t ready yet get follow-ups that match what they’ve read.

Total183h/mo$143k/yr

What changes

Each demo request lands with the right rep, a company brief and a drafted reply while the prospect is still paying attention.

  • 3steps automated
  • 5steps for an AI agent
  • 2calls kept with a person

What if you had it

183 hours a month back$142,700 a year in time and savings

For a sales team of about 12 reps handling around 300 inbound demo requests a month, at $65 an hour.

Customize to your business

Pricing and quoting

Quotes priced right and approved before they go out

Starts with

Rep asks for a quote

A rep needs a quote for a new deal or a renewal.

Workflow automation: Brings the account’s product usage, its segment, the current price book and any running promotion onto the deal.
What if: 40 hours a month back, worth $33,600 a year

What if you had it

40 hours a month back$33,600 a year

About 20 minutes gathering usage and prices for each of 120 quotes.

Customize to your business
Starts when
A rep needs a quote for a new deal or a renewal.
Result
Everything the quote needs is on the deal.

Why each part sits where it does

  • Workflow automation. The same four facts come from the same four places on every deal.

Reads

  • CRM
  • Product usage
  • Price book

Writes

  • CRM
Step 1 of 5
AI agent: Reads the rep’s note (“40 seats, three sites, starting in March”) and works out the products, quantities and start date it means.
Workflow automation: Prices each line from the price book, applies volume breaks and the promotion, and totals it with your standard payment terms.
What if: 50 hours a month back, worth $42,000 a year

What if you had it

50 hours a month back$42,000 a year

About 25 minutes building line items on each of 120 quotes.

Customize to your business
Starts when
Everything the quote needs is on the deal.
Result
A priced draft that matches what the rep asked for.

Why each part sits where it does

  • AI agent. Reps ask for quotes in shorthand, and “three sites” has to become three site licenses.
  • Workflow automation. List prices and volume breaks are written down. After that it’s multiplication.

Reads

  • CRM
  • Price book

Writes

  • CRM

The agent

Quote creatorDrafts quotes from your price book
Step 2 of 5
Workflow automation: Stops any quote discounted past the 15% a rep can give on their own, and sends it to the sales director.
AI agent: Pulls recent deals of the same size and segment from the CRM, what each was discounted and whether it closed, and sums it up in two lines.
If over 15%Person: Approves the 18% with those past deals beside it, or sends the rep a counter at 12%.Declined: back to the rep with a counter
What if: 15 hours a month back, worth $42,600 a year

What if you had it

15 hours a month back$42,600 a year

About 30 minutes finding past deals for each of 30 held quotes, plus $30,000 in discount margin kept.

Customize to your business
Starts when
A priced draft that matches what the rep asked for.
Result
No discount goes out that nobody approved.

Why each part sits where it does

  • Workflow automation. A 15% limit is a line anyone can check. The quote stops the moment it’s crossed.
  • AI agent. Which past deals count as “like this one” depends on size, segment and timing at once.
  • Person. Every point of discount comes out of margin, and above 15% that trade is the director’s to make.

Reads

  • Discount rules
  • CRM

Writes

  • CRM
  • Chat

The agent

Ask the pipelineAnswers pipeline questions from your CRM
Step 3 of 5
AI agent: Writes the proposal from the discovery call and the approved quote: their goals in their words, the scope, a matching case study, a timeline.
If new scopePerson: Decides whether work you haven’t sold before, like a custom integration, stays in the proposal or comes out.Declined: cut from the proposal
What if: 60 hours a month back, worth $50,400 a year

What if you had it

60 hours a month back$50,400 a year

About 90 minutes drafting each of 40 proposals.

Customize to your business
Starts when
No discount goes out that nobody approved.
Result
A proposal that agrees with the quote, line for line.

Why each part sits where it does

  • AI agent. A proposal lands when it plays back what the client said on the call, in their words.
  • Person. Someone has to be sure the team can deliver it before it’s promised in writing.

Reads

  • Recordings
  • CRM

Writes

  • Shared drive

The agent

Proposal generatorTurns a discovery call into a proposal
Step 4 of 5
Workflow automation: Sends both for e-signature, files them on the deal and gives the rep a follow-up task for day five.E-signDeal recordRep task
What if: 12 hours a month back, worth $10,100 a year

What if you had it

12 hours a month back$10,100 a year

About 6 minutes sending and filing each of 120 quotes.

Customize to your business
Starts when
A proposal that agrees with the quote, line for line.
Result
The follow-up is on the rep’s list before the client opens the quote.

Why each part sits where it does

  • Workflow automation. Sending, filing and a day-five reminder never change from one deal to the next.

Reads

  • CRM

Writes

  • E-sign
  • CRM
  • Email
Step 5 of 5

Ends with

Out for signature

The client has a priced, approved quote to sign, it’s logged on the deal, and the rep’s follow-up is already booked.

Total177h/mo$179k/yr

What changes

Reps get quotes back within hours, priced from the price book. Any discount past the limit reaches the director before the client.

  • 4steps automated
  • 3steps for an AI agent
  • 2calls kept with a person

What if you had it

177 hours a month back$178,700 a year in time and savings

For a sales team sending about 120 quotes and renewals a month, 40 of them with a proposal, at $70 an hour.

Customize to your business

One recording, a month of posts

Turning a podcast or webinar into clips and posts

Starts with

Recording lands

A finished podcast episode or webinar is saved to the shared drive.

Workflow automation: Picks up the new file, transcribes it with speaker names and timestamps, and saves the transcript beside the recording.
What if: 8 hours a month back, worth $5,600 a year

What if you had it

8 hours a month back$5,600 a year

About 90 minutes transcribing and cleaning up each of 5 recordings.

Customize to your business
Starts when
A finished podcast episode or webinar is saved to the shared drive.
Result
A searchable transcript within the hour.

Why each part sits where it does

  • Workflow automation. Speech-to-text is a standard tool now, and it runs on every file without anyone listening.

Reads

  • Shared drive

Writes

  • Recordings
Step 1 of 5
AI agent: Marks eight to ten moments that make sense without the rest of the episode, each with timestamps, a working title and why it works.
Person: Keeps the moments that sound like your company, drops the rest and adds any the agent missed.
What if: 15 hours a month back, worth $10,400 a year

What if you had it

15 hours a month back$10,400 a year

About 3 hours listening back to each of 5 recordings.

Customize to your business
Starts when
A searchable transcript within the hour.
Result
A shortlist of clips, chosen by a person.

Why each part sits where it does

  • AI agent. Finding the minute in an hour that works on its own used to mean an afternoon of listening.
  • Person. What the company sounds like in public is a marketer’s call.

Reads

  • Recordings

Writes

  • Project tool

The agent

Recordings into clipsTurns podcasts and webinars into shareable cuts
Step 2 of 5
Workflow automation: Cuts each chosen moment at its timestamps, adds captions from the transcript and saves every clip to one folder.
What if: 15 hours a month back, worth $28,400 a year

What if you had it

15 hours a month back$28,400 a year

About 20 minutes cutting and captioning each of 45 clips, plus $18,000 in freelance video editing not booked.

Customize to your business
Starts when
A shortlist of clips, chosen by a person.
Result
A folder of captioned clips, ready to post.

Why each part sits where it does

  • Workflow automation. Once the timestamps are approved, cutting and captioning is mechanical.

Reads

  • Recordings
  • Project tool

Writes

  • Shared drive
Step 3 of 5
Workflow automation: Checks competitor sites, their social accounts and your news alerts at 7:00 each morning, and collects anything new in one list.
AI agent: Sorts the week’s news, drops the noise and flags the stories where one of your clips has something to add.
What if: 11 hours a month back, worth $7,700 a year

What if you had it

11 hours a month back$7,700 a year

About 30 minutes checking the news each of 21 workdays.

Customize to your business
Starts when
A folder of captioned clips, ready to post.
Result
Each post has a reason to go out this week.

Why each part sits where it does

  • Workflow automation. The sources are a fixed list, checked at the same time every day.
  • AI agent. In a subject line, the one story that matters looks like the forty that don’t.

Reads

  • News feeds

Writes

  • Project tool
  • Email

The agent

Competitor and market digestReads the news and sends what matters
Step 4 of 5
AI agent: Writes a post for each clip in your house style, ties it to a news story where one fits and spaces them over four weeks.
Person: Reads every post before it goes out, and cuts any line that sounds like someone else’s company.Declined: back to the agent with notes
Workflow automation: Schedules each approved post on its channel and date, and sends the guest their clips to share.Content calendarSocial accountsGuest email
What if: 38 hours a month back, worth $26,400 a year

What if you had it

38 hours a month back$26,400 a year

About 25 minutes writing and scheduling each of 90 posts.

Customize to your business
Starts when
Each post has a reason to go out this week.
Result
A month of posts on the calendar, each one read by a person.

Why each part sits where it does

  • AI agent. Twenty posts from one recording still need to read like twenty different posts.
  • Person. Once a post is live it speaks for the company, and a person should have read it first.
  • Workflow automation. A scheduled post needs a time and a channel, and both are already set.

Reads

  • Shared drive
  • Project tool

Writes

  • Content calendar
  • Social accounts
  • Email

The agent

Recordings into clipsTurns podcasts and webinars into shareable cuts
Step 5 of 5

Ends with

A month on the calendar

Captioned clips and approved posts are scheduled across the next four weeks.

Total87h/mo$78k/yr

What changes

One webinar becomes a month of clips and posts. Your marketer’s part shrinks to choosing the moments and approving what goes out.

  • 4steps automated
  • 3steps for an AI agent
  • 2calls kept with a person

What if you had it

87 hours a month back$78,500 a year in time and savings

For a marketing team turning about 5 podcast episodes and webinars a month into clips and posts, at $58 an hour.

Customize to your business

Operations AI for Operations →

Software vendor approval

Approving a new software tool someone asked for

Starts with

Tool requested

Someone drops a link in chat and asks whether the team can get it.

AI agent: Works out what was actually asked for from the message and the vendor’s pricing page: the purpose, 12 seats, about $9,000 a year, the data it touches.
Workflow automation: Adds the request to the software list and replies in the thread with its number, so the requester can see where it stands.
What if: 15 hours a month back, worth $11,700 a year

What if you had it

15 hours a month back$11,700 a year

About 90 minutes of back-and-forth on each of 10 requests.

Customize to your business
Starts when
Someone drops a link in chat and asks whether the team can get it.
Result
A complete request, with the cost, seats and data written down.

Why each part sits where it does

  • AI agent. The request is a sentence and a link. Seats, cost and the data it holds are somewhere on the vendor’s site.
  • Workflow automation. Adding a row and posting its number back is bookkeeping.

Reads

  • Chat

Writes

  • Software list
  • Chat

The agent

Software vendor approvalRoutes new tools past IT, legal and finance
Step 1 of 6
AI agent: Checks the request against every tool on the software list and flags one that does the same job, like the survey tool with 40 unused seats.
If you already have oneWorkflow automation: Replies with the existing tool, who owns it and how to get a seat, then closes the request.
What if: 8 hours a month back, worth $24,200 a year

What if you had it

8 hours a month back$24,200 a year

About 45 minutes checking the software list on each of 10 requests, plus $18,000 in duplicate tools never bought.

Customize to your business
Starts when
A complete request, with the cost, seats and data written down.
Result
Either a genuinely new need, or a seat on a tool you own.

Why each part sits where it does

  • AI agent. Two tools can do the same job under completely different names. Spotting it means knowing what each one is for.
  • Workflow automation. Once the match is made, the reply is a fill-in-the-blanks message.

Reads

  • Software list

Writes

  • Software list
  • Chat

The agent

Software vendor approvalRoutes new tools past IT, legal and finance
Step 2 of 6
AI agent: Goes through the vendor’s security report, privacy terms and insurance certificate, then lists what they’d hold, where it’s stored and what’s missing.
If it touches customer dataPerson: Reads the review and decides whether the tool can hold customer data, or sends the vendor a list of questions first.Declined: back to the vendor for answers
What if: 20 hours a month back, worth $15,600 a year

What if you had it

20 hours a month back$15,600 a year

About 4 hours on each of 5 vendors that touch customer data.

Customize to your business
Starts when
Either a genuinely new need, or a seat on a tool you own.
Result
The security review is done and on file before anything is signed.

Why each part sits where it does

  • AI agent. Every vendor answers security questions in its own format, from a two-page letter to a 90-page report.
  • Person. Letting an outside company hold customer data is a risk someone inside the company has to own.

Reads

  • Email
  • Shared drive

Writes

  • Shared drive

The agent

Vendor risk reviewsRuns security checks on every new vendor
Step 3 of 6
AI agent: Finds the auto-renewal, the cap on price increases, the liability limit and the data terms, and marks them for the reviewer.
If over $50k a yearWorkflow automation: Adds legal to the approvers when the contract is worth more than $50,000 a year, with the marked clauses attached.
What if: 10 hours a month back, worth $7,800 a year

What if you had it

10 hours a month back$7,800 a year

About 2 hours reading each of 5 contracts for the key clauses.

Customize to your business
Starts when
The security review is done and on file before anything is signed.
Result
The four clauses worth reading are marked, and legal is in when the size calls for it.

Why each part sits where it does

  • AI agent. Renewal terms turn up in a different section of every contract, often under a different heading.
  • Workflow automation. The $50,000 line is in your approval policy and the value is on the order form. Comparing two numbers is a rule.

Reads

  • Contracts

Writes

  • Software list

The agent

Software vendor approvalRoutes new tools past IT, legal and finance
Step 4 of 6
Workflow automation: Sends the request, review and marked contract to every approver together and nudges anyone quiet for two days.FinanceDepartment headLegal, over $50k
Person: Finance, the department head and, over $50,000, legal each approve. The request moves on only once all of them have.Declined: back to the requester with the reason
What if: 10 hours a month back, worth $7,800 a year

What if you had it

10 hours a month back$7,800 a year

About 1 hour of chasing approvers on each of 10 requests.

Customize to your business
Starts when
The four clauses worth reading are marked, and legal is in when the size calls for it.
Result
Everyone who had to agree has, and it’s on the record.

Why each part sits where it does

  • Workflow automation. Who approves what comes from your policy, and a reminder only needs a date.
  • Person. Whether the spend is worth it is the budget owner’s call. Whether the tool fits the team is the department head’s.

Reads

  • Software list

Writes

  • Software list
  • Email
Step 5 of 6
Workflow automation: Files the contract and review in the vendor’s folder, raises the PO and logs the renewal date.Vendor folderPurchase orderRenewal date
What if: 5 hours a month back, worth $13,900 a year

What if you had it

5 hours a month back$13,900 a year

About 45 minutes filing and raising the PO on each of 7 approvals, plus $10,000 in renewals cancelled before they bill.

Customize to your business
Starts when
Everyone who had to agree has, and it’s on the record.
Result
The tool is paid for, filed, and on the list well before it renews.

Why each part sits where it does

  • Workflow automation. After the last approval, the paperwork is identical for every tool.

Reads

  • Software list

Writes

  • Shared drive
  • Accounting
  • Software list
Step 6 of 6

Ends with

Approved and filed

Every sign-off it needed is on record, the PO is raised and the contract sits in the vendor’s folder.

Total68h/mo$81k/yr

What changes

Anyone who asks for a tool gets an answer in days, or a seat on the one you already pay for.

  • 5steps automated
  • 4steps for an AI agent
  • 2calls kept with a person

What if you had it

68 hours a month back$81,000 a year in time and savings

For a 300-person company fielding about 10 new software requests a month, at $65 an hour.

Customize to your business

Weekly KPI roll-up

Every team’s numbers in one Friday report

Starts with

Friday, 2:00 p.m.

The week closes with each team’s numbers still sitting in its own system.

Workflow automation: Collects the week’s figures from sales, finance, HR, support and delivery into one KPI sheet.CRMAccountingHR systemTicket queue
What if: 43 hours a month back, worth $36,100 a year

What if you had it

43 hours a month back$36,100 a year

About 2 hours a week from each of 5 teams, pulling its figures.

Customize to your business
Starts when
The week closes with each team’s numbers still sitting in its own system.
Result
This week’s raw numbers, together in one sheet.

Why each part sits where it does

  • Workflow automation. Each Friday it’s five systems and a known list of fields.

Reads

  • CRM
  • Accounting
  • HR system
  • Ticket queue
  • Project tool

Writes

  • KPI sheet
Step 1 of 5
AI agent: Maps every team’s figures to the agreed definitions, so “active customers” counts the same accounts in the sales numbers and the support numbers.
If a team’s numbers are missingWorkflow automation: Messages the owner of any missing figure in chat and drops their answer into the sheet when they reply.
What if: 17 hours a month back, worth $14,300 a year

What if you had it

17 hours a month back$14,300 a year

About 4 hours a week matching definitions and chasing late teams.

Customize to your business
Starts when
This week’s raw numbers, together in one sheet.
Result
One set of numbers that means the same thing on every team.

Why each part sits where it does

  • AI agent. Teams label things their own way. Sales may count active customers by contract, support by who opened a ticket this month.
  • Workflow automation. A blank cell with an owner’s name beside it is all a reminder needs.

Reads

  • KPI sheet

Writes

  • KPI sheet
  • Chat

The agent

KPI roll-upPulls every team’s numbers into one view
Step 2 of 5
Workflow automation: Marks every KPI more than 2 points off plan, using the range leadership agreed in January.
AI agent: Looks at the orders, tickets and projects behind a flagged number and writes two lines on what changed, like one $40,000 invoice that slipped.
What if: 26 hours a month back, worth $21,800 a year

What if you had it

26 hours a month back$21,800 a year

About 45 minutes on each of 8 misses a week.

Customize to your business
Starts when
One set of numbers that means the same thing on every team.
Result
A reason written beside each miss.

Why each part sits where it does

  • Workflow automation. Checking whether a number falls outside a range is arithmetic.
  • AI agent. A dip in margin could be one late invoice or a price change nobody mentioned, and the dashboard shows both as the same red number.

Reads

  • KPI sheet
  • CRM
  • Project tool

Writes

  • KPI sheet

The agent

The Friday reportHas the weekly report ready by Friday
Step 3 of 5
AI agent: Writes it in the format leadership already reads: headline numbers first, then what moved, then a short note on each miss.
If anything is under planPerson: Rewrites any note on a miss that needs more context, then approves the report to go out.Declined: back to the agent with edits
What if: 22 hours a month back, worth $18,500 a year

What if you had it

22 hours a month back$18,500 a year

About 5 hours a week drafting and prepping the COO review.

Customize to your business
Starts when
A reason written beside each miss.
Result
A report leadership can get through in five minutes.

Why each part sits where it does

  • AI agent. Turning a table into three paragraphs means choosing what mattered this week and what can wait.
  • Person. The explanation of a miss goes in front of leadership under the COO’s name.

Reads

  • KPI sheet

Writes

  • Shared drive

The agent

The Friday reportHas the weekly report ready by Friday
Step 4 of 5
Workflow automation: Refreshes the dashboard, emails the leadership list and posts each team’s numbers in its own channel.DashboardLeadership emailTeam channels
What if: 7 hours a month back, worth $5,900 a year

What if you had it

7 hours a month back$5,900 a year

About 90 minutes a week on the dashboard, email and channels.

Customize to your business
Starts when
A report leadership can get through in five minutes.
Result
Everyone starts Monday from the same numbers.

Why each part sits where it does

  • Workflow automation. Who gets the report was decided once, and the channels already exist.

Reads

  • KPI sheet

Writes

  • Dashboard
  • Email
  • Chat
Step 5 of 5

Ends with

Report out by Friday

The dashboard is current and leadership has the week’s numbers, with a note on anything that missed plan.

Total115h/mo$97k/yr

What changes

Leadership reads the week’s numbers on Friday afternoon, with a reason already written beside every miss.

  • 4steps automated
  • 3steps for an AI agent
  • 1call kept with a person

What if you had it

115 hours a month back$96,600 a year in time and savings

For a 300-person company with five teams reporting their numbers every week, at $70 an hour.

Customize to your business

Signed order to every system

Setting up a new customer once the order is signed

Starts with

Order signed

The customer signs the order form and the deal is marked won.

Workflow automation: Grabs the order from e-sign as soon as the last signature lands and marks the deal won in the CRM.
AI agent: Combines the order form, the quote and the rep’s call notes into one summary: products, start date, contacts, discount and anything promised.
What if: 30 hours a month back, worth $21,600 a year

What if you had it

30 hours a month back$21,600 a year

About 45 minutes piecing together each of 40 orders from 3 documents.

Customize to your business
Starts when
The customer signs the order form and the deal is marked won.
Result
One summary of the deal that delivery and finance both work from.

Why each part sits where it does

  • Workflow automation. E-sign reports the last signature the moment it lands.
  • AI agent. What was sold is spread across three documents, and a line like “we’ll waive the setup fee” may only be in the rep’s notes.

Reads

  • E-sign
  • CRM

Writes

  • CRM

The agent

Cross-system handoffsSets up a signed order in every system
Step 1 of 5
AI agent: Checks the summary for anything off the price book: free months, a side deal in the notes, an unusual start date, a discount past the rep’s limit.
If over the discount limitPerson: Approves the billing schedule when the discount is over the rep’s 15% limit, or sends the order back to sales to fix.Declined: back to the rep to fix the order
What if: 20 hours a month back, worth $32,400 a year

What if you had it

20 hours a month back$32,400 a year

About 30 minutes checking terms on each of 40 orders, plus $18,000 in setup fees and extras now billed.

Customize to your business
Starts when
One summary of the deal that delivery and finance both work from.
Result
Terms finance can invoice without a phone call.

Why each part sits where it does

  • AI agent. A side deal can be one line in a call note. It still changes what finance is allowed to invoice.
  • Person. Past the discount a rep may give, someone who owns the numbers has to agree.

Reads

  • CRM
  • Price book

Writes

  • CRM

The agent

Cross-system handoffsSets up a signed order in every system
Step 2 of 5
Workflow automation: Sets up the project, the customer in accounting and the billing schedule from the one approved summary.ProjectCustomer in accountingBilling schedule
What if: 40 hours a month back, worth $28,800 a year

What if you had it

40 hours a month back$28,800 a year

About 1 hour keying each of 40 customers into three systems.

Customize to your business
Starts when
Terms finance can invoice without a phone call.
Result
The same customer, spelled the same way, in every system.

Why each part sits where it does

  • Workflow automation. With the details checked, entering them in three systems is copying.

Reads

  • CRM

Writes

  • Project tool
  • Accounting
Step 3 of 5
AI agent: Builds a setup checklist from what this customer bought and the last similar account, then drafts a welcome email asking for the first three things.
Person: Checks the checklist, picks who’s on the account and rewrites the welcome email wherever it doesn’t sound like the team.Declined: back to the agent with changes
What if: 60 hours a month back, worth $43,200 a year

What if you had it

60 hours a month back$43,200 a year

About 90 minutes on the plan and welcome for each of 40 customers.

Customize to your business
Starts when
The same customer, spelled the same way, in every system.
Result
A checklist, a team and a welcome email, ready before kickoff.

Why each part sits where it does

  • AI agent. Every customer buys a slightly different mix, and the checklist has to match what this one bought.
  • Person. Staffing depends on who has room this month. The delivery lead knows that and the system doesn’t.

Reads

  • CRM
  • Project tool

Writes

  • Project tool

The agent

Customer onboardingRuns each new account’s setup checklist
Step 4 of 5
Workflow automation: Books the kickoff call, sends the welcome and the first invoice, files the order and posts the win in chat.Kickoff inviteFirst invoiceWin posted
What if: 27 hours a month back, worth $19,400 a year

What if you had it

27 hours a month back$19,400 a year

About 40 minutes booking kickoff and invoicing each of 40.

Customize to your business
Starts when
A checklist, a team and a welcome email, ready before kickoff.
Result
The customer has a kickoff date and finance has an invoice out.

Why each part sits where it does

  • Workflow automation. Kickoff, welcome, invoice and the win post happen for every customer, in that order.

Reads

  • Project tool
  • Accounting

Writes

  • Calendars
  • Email
  • Accounting
  • Shared drive
  • Chat
Step 5 of 5

Ends with

Kicked off and billed

The project is set up, kickoff is on the calendar, the first invoice is out and every team knows.

Total177h/mo$145k/yr

What changes

Delivery and finance start from one summary of the deal, including whatever the rep promised on the call.

  • 3steps automated
  • 3steps for an AI agent
  • 2calls kept with a person

What if you had it

177 hours a month back$145,400 a year in time and savings

For a company of about 300 people signing around 40 new customer orders a month, at $60 an hour.

Customize to your business

HR AI for HR →

New hire onboarding

Everything a new hire needs before they start

Starts with

Offer signed

The candidate signs the offer in applicant tracking.

AI agent: Compares the title, salary and start date on the signed offer with the band for the role. An $88,000 offer on an $80,000 ceiling gets flagged.
If outside the bandPerson: Looks at any salary over the band and approves it or sends it back, before it goes anywhere near payroll.Declined: back to the recruiter
What if: 10 hours a month back, worth $16,000 a year

What if you had it

10 hours a month back$16,000 a year

About 40 minutes of band checks on each of 15 offers, plus $9,000 in above-band pay caught before payroll.

Customize to your business
Starts when
The candidate signs the offer in applicant tracking.
Result
The employee record is created once, with the right numbers.

Why each part sits where it does

  • AI agent. Recruiters write offers their own way, and the salary can sit in a table, a sentence or a side letter.
  • Person. Paying above the band is a judgment about this hire, and HR owns it.

Reads

  • Applicant tracking
  • Pay bands

Writes

  • HR system

The agent

Offer letter checksChecks every offer against the pay band
Step 1 of 5
Workflow automation: Sends a single secure link for bank details, tax forms and an emergency contact, and files the answers straight to the HR system.
AI agent: Notices the W-4 left unsigned or the bank details missing a routing number, and asks the new hire for that one thing.
What if: 38 hours a month back, worth $26,400 a year

What if you had it

38 hours a month back$26,400 a year

About 2.5 hours of form chasing on each of 15 new hires.

Customize to your business
Starts when
The employee record is created once, with the right numbers.
Result
Tax, bank and I-9 forms are complete before the start date.

Why each part sits where it does

  • Workflow automation. The form, and where it gets filed, are the same for every hire.
  • AI agent. A form can come back full and still be wrong. Catching that means actually reading it.

Reads

  • HR system

Writes

  • HR system
  • E-sign

The agent

Compliance paperworkFiles and tracks the forms every hire needs
Step 2 of 5
AI agent: Looks at the role, the team and what the last person in the job had, then picks the systems, folders and groups to open on day one.
Workflow automation: Opens each account on the list and orders the laptop to arrive the day before they start.Email and chatPayrollLaptop order
What if: 45 hours a month back, worth $31,300 a year

What if you had it

45 hours a month back$31,300 a year

About 3 hours of access lists and IT tickets on each of 15 hires.

Customize to your business
Starts when
Tax, bank and I-9 forms are complete before the start date.
Result
Logins made, and the laptop on its way.

Why each part sits where it does

  • AI agent. Two people with the same title can need very different access, depending on the team and the clients they cover.
  • Workflow automation. With the list decided, there’s nothing left to judge.

Reads

  • HR system
  • IT access

Writes

  • IT tickets
  • Payroll
  • Email

The agent

New hire onboardingTurns a signed offer into day one
Step 3 of 5
AI agent: Writes a first-month plan from the job description, the team’s current projects and what the last hire in this role worked on.
Person: Changes what they want, adds the people worth meeting, and approves it.Declined: back to the agent with notes
What if: 30 hours a month back, worth $20,900 a year

What if you had it

30 hours a month back$20,900 a year

About 2 hours of plan writing saved on each of 15 hires.

Customize to your business
Starts when
Logins made, and the laptop on its way.
Result
The new hire has a plan to read on their first morning.

Why each part sits where it does

  • AI agent. Most managers mean to write this plan and run out of week. A draft gets it written.
  • Person. Only the manager knows what they hired this person to fix.

Reads

  • HR system
  • Project tool

Writes

  • Project tool

The agent

New hire onboardingTurns a signed offer into day one
Step 4 of 5
Workflow automation: Puts the first week’s meetings on calendars, sends the welcome email and tells the team who’s starting.CalendarsWelcome emailTeam notice
What if: 23 hours a month back, worth $16,000 a year

What if you had it

23 hours a month back$16,000 a year

About 1.5 hours of booking and notices on each of 15 hires.

Customize to your business
Starts when
The new hire has a plan to read on their first morning.
Result
The first week is booked and the team knows who’s coming.

Why each part sits where it does

  • Workflow automation. Every step here follows from the start date. Plain automation handles it well.

Reads

  • Calendars

Writes

  • Calendars
  • Email
  • Chat
Step 5 of 5

Ends with

Ready on day one

The laptop works, the logins exist, and the first week is already on their calendar.

Total146h/mo$111k/yr

What changes

New hires walk in to a working laptop, every login they need and a first week already on the calendar.

  • 3steps automated
  • 4steps for an AI agent
  • 2calls kept with a person

What if you had it

146 hours a month back$110,600 a year in time and savings

For a 450-person company bringing on about 15 new hires a month, at $58 an hour.

Customize to your business

Time off and leave

Leave requests handled in the chat where people ask

Starts with

Time off requested

Someone asks for days off in chat, in their own words.

AI agent: Turns “can I take Friday and Monday?” into two dates, checks them against the balance and the carry-over rule, and replies in the thread.
Workflow automation: Records the request on the employee’s file, so there’s one list of who asked for what, and when.
What if: 40 hours a month back, worth $30,500 a year

What if you had it

40 hours a month back$30,500 a year

About 8 minutes to read and log each of 300 requests, plus $7,500 in leave paid on balances that were wrong.

Customize to your business
Starts when
Someone asks for days off in chat, in their own words.
Result
A request with real dates and enough balance to cover it.

Why each part sits where it does

  • AI agent. People ask in sentences. The HR system wants two dates and a leave type.
  • Workflow automation. Once the dates are clear, logging them is a form fill.

Reads

  • Chat
  • Leave balances
  • Handbook

Writes

  • HR system

The agent

Policy and leave questionsAnswers HR questions from the handbook
Step 1 of 5
AI agent: Looks at the team calendar for those dates and flags the request if the team would be down to two people on the Friday.
What if: 25 hours a month back, worth $14,400 a year

What if you had it

25 hours a month back$14,400 a year

About 5 minutes of calendar checks on each of 300 requests.

Customize to your business
Starts when
A request with real dates and enough balance to cover it.
Result
A clean request, or a flag that says exactly where the gap is.

Why each part sits where it does

  • AI agent. Three people out can be fine for one team in August and a real problem for another at month-end.

Reads

  • Calendars
  • HR system

Writes

  • HR system

The agent

Leave requestsApproves time off, updates calendars
Step 2 of 5
Workflow automation: Sends the request to the employee’s team lead, with the balance and any clash attached.
Person: Says yes to the days, or suggests another week if that Friday leaves the team too thin.Declined: back to the employee with the reason
What if: 20 hours a month back, worth $11,500 a year

What if you had it

20 hours a month back$11,500 a year

About 4 minutes of looking up balance and clashes, on 300 requests.

Customize to your business
Starts when
A clean request, or a flag that says exactly where the gap is.
Result
The employee gets a yes or no in the same thread.

Why each part sits where it does

  • Workflow automation. Who approves whom is already in the org chart.
  • Person. The lead knows what’s due that week, and whether the team can spare someone.

Reads

  • Org chart

Writes

  • HR system
  • Chat
Step 3 of 5
Workflow automation: Puts the leave on the team and manager calendars, takes the days off the balance and tells the project lead.Team calendarBalanceProject lead
What if: 30 hours a month back, worth $17,300 a year

What if you had it

30 hours a month back$17,300 a year

About 6 minutes of calendar and balance updates on each of 300.

Customize to your business
Starts when
The employee gets a yes or no in the same thread.
Result
Nobody finds out on the morning they’re out.

Why each part sits where it does

  • Workflow automation. An approval is a yes in a field. Everything after it follows from that.

Reads

  • HR system

Writes

  • Calendars
  • Leave balances
  • Email
Step 4 of 5
Workflow automation: Posts a note in the team channel seven days before the leave starts.
AI agent: Pulls the person’s open tasks, picks out the three that will move while they’re away, and drafts a note for them to check and send.
What if: 33 hours a month back, worth $19,000 a year

What if you had it

33 hours a month back$19,000 a year

About 20 minutes of handover notes on each of 100 longer absences.

Customize to your business
Starts when
Nobody finds out on the morning they’re out.
Result
The team knows, and the work that can’t wait has an owner.

Why each part sits where it does

  • Workflow automation. A reminder on a fixed date is a simple rule.
  • AI agent. Most open tasks can wait a week. The three that can’t rarely say so in their titles.

Reads

  • Project tool

Writes

  • Chat

The agent

Leave requestsApproves time off, updates calendars
Step 5 of 5

Ends with

Approved and covered

The leave is on every calendar, the balance is right, and the team gets a reminder a week out.

Total148h/mo$93k/yr

What changes

People ask for days off the way they’d ask a colleague. Their team lead sees the balance and who else is out before saying yes.

  • 4steps automated
  • 3steps for an AI agent
  • 1call kept with a person

What if you had it

148 hours a month back$92,700 a year in time and savings

For a 450-person company handling about 300 leave requests a month, at $48 an hour.

Customize to your business

Offboarding

The last day: access closed, final pay right

Starts with

Last day set

HR saves a last day on the employee’s record.

Workflow automation: Starts the offboarding checklist as soon as the last day is saved, and assigns each item to HR, IT or the manager.
AI agent: Goes through access records, shared drives and group lists, down to the old project folder and the vendor portal nobody remembers granting.
What if: 20 hours a month back, worth $13,200 a year

What if you had it

20 hours a month back$13,200 a year

About 2.5 hours of account hunting on each of 8 leavers.

Customize to your business
Starts when
HR saves a last day on the employee’s record.
Result
One list of every account to close.

Why each part sits where it does

  • Workflow automation. A saved last day is an event the HR system already announces.
  • AI agent. Access piles up over years and gets recorded in five different places, when it gets recorded at all.

Reads

  • HR system
  • IT access

Writes

  • IT tickets

The agent

OffboardingCloses every account on the last day
Step 1 of 5
AI agent: Lists their open projects, clients and files, and suggests who takes each one, starting with whoever already works with that client.
What if: 24 hours a month back, worth $15,800 a year

What if you had it

24 hours a month back$15,800 a year

About 3 hours of handover writing on each of 8 leavers.

Customize to your business
Starts when
One list of every account to close.
Result
Open work has a new owner before they go.

Why each part sits where it does

  • AI agent. The manager edits a draft instead of piecing it together during the leaver’s last week.

Reads

  • Project tool
  • Shared drive

Writes

  • Project tool

The agent

OffboardingCloses every account on the last day
Step 2 of 5
AI agent: Adds up the last paycheck from the contract and the leave balance: 3.5 unused days paid out, less a $600 advance still owed.
Person: Checks the figures and confirms them before anything goes to payroll.Declined: back with the correction
What if: 12 hours a month back, worth $14,900 a year

What if you had it

12 hours a month back$14,900 a year

About 1.5 hours of final-pay math on each of 8 leavers, plus $7,000 in final-pay errors caught before they run.

Customize to your business
Starts when
Open work has a new owner before they go.
Result
The last paycheck is right the first time.

Why each part sits where it does

  • AI agent. Notice periods, leave rules and what’s owed back differ from one contract to the next.
  • Person. A final paycheck is hard to fix once it’s paid, so a person checks it first.

Reads

  • Leave balances
  • Payroll

Writes

  • Payroll

The agent

OffboardingCloses every account on the last day
Step 3 of 5
Workflow automation: At 6pm on the last day all three happen at once, so nothing waits for someone to remember.Access removedLaptop returnFiles moved
What if: 12 hours a month back, worth $20,400 a year

What if you had it

12 hours a month back$20,400 a year

About 1.5 hours of closing access and chasing laptops, 8 leavers, plus $12,500 in software seats cancelled on the last day.

Customize to your business
Starts when
The last paycheck is right the first time.
Result
Nothing is left open once they’ve gone.

Why each part sits where it does

  • Workflow automation. This one is about timing. A scheduled rule runs at the right minute every time.

Reads

  • IT access

Writes

  • IT access
  • IT tickets
  • Shared drive
Step 4 of 5
AI agent: Compares the HR system, payroll and the org chart, and flags the one still showing them on the sales team.
Workflow automation: Moves their direct reports to the new manager and takes their box off the org chart.
What if: 12 hours a month back, worth $7,900 a year

What if you had it

12 hours a month back$7,900 a year

About 1.5 hours of cross-checking three systems on each of 8.

Customize to your business
Starts when
Nothing is left open once they’ve gone.
Result
Headcount matches in every system.

Why each part sits where it does

  • AI agent. Systems drift apart quietly. Finding where takes comparing them line by line.
  • Workflow automation. Once the new manager is named, moving the reports is a simple edit.

Reads

  • HR system
  • Payroll
  • Org chart

Writes

  • Org chart

The agent

Headcount reconciliationKeeps HR, payroll and the org chart agreeing
Step 5 of 5

Ends with

Closed on the last day

Access off, the laptop on its way back, final pay right, and every system showing the same date.

Total80h/mo$72k/yr

What changes

Access closes on the last day, down to the vendor portal nobody remembered granting, with final pay checked before it runs.

  • 3steps automated
  • 4steps for an AI agent
  • 1call kept with a person

What if you had it

80 hours a month back$72,200 a year in time and savings

For a 450-person company with about 8 people leaving a month, at $55 an hour.

Customize to your business

Customer Support AI for Customer Support →

Ticket to resolution

Ticket answers that stay inside your refund policy

Starts with

Customer writes in

An email or chat arrives, often with a photo of what went wrong.

Workflow automation: Matches the sender’s email to their account and pins the tier and last five orders to the ticket as it lands.
AI agent: Looks at the message and the photo of a split carton, decides it’s damage in transit on order #40219, and tags the ticket.
What if: 60 hours a month back, worth $30,200 a year

What if you had it

60 hours a month back$30,200 a year

About 2 minutes on each of 1,800 tickets.

Customize to your business
Starts when
An email or chat arrives, often with a photo of what went wrong.
Result
The ticket carries its order, account and problem.

Why each part sits where it does

  • Workflow automation. Matching an email address to an account is a lookup.
  • AI agent. One customer writes “arrived broken”, another sends a photo and no words. Both are the same fault.

Reads

  • Ticket queue
  • CRM
  • Orders

Writes

  • Ticket queue

The agent

Ticket triage and refundsResolves what is within your limits
Step 1 of 5
If they ask for a personAI agent: Sums up the thread in four lines: what went wrong, what’s been tried, what the customer was told and what they want now.
Person: Opens the ticket to a written summary and replies without asking the customer to explain it all again.
What if: 30 hours a month back, worth $15,100 a year

What if you had it

30 hours a month back$15,100 a year

About 10 minutes of catching up on each of 180 handovers.

Customize to your business
Starts when
The ticket carries its order, account and problem.
Result
The ticket changes hands with the story written down.

Why each part sits where it does

  • AI agent. The useful history can be spread across a dozen messages and two channels.
  • Person. Anyone who asks for a person gets one.

Reads

  • Ticket queue
  • CRM
  • Orders

Writes

  • Ticket queue

The agent

Escalation handoffHands tickets over with the full story
Step 2 of 5
AI agent: Finds what your policy allows for transit damage on a Standard account, proposes a replacement or refund, and drafts the reply.
What if: 45 hours a month back, worth $37,700 a year

What if you had it

45 hours a month back$37,700 a year

About 3 minutes on each of 900 replies that need a fix, plus $15,000 in refunds kept inside your policy.

Customize to your business
Starts when
The ticket changes hands with the story written down.
Result
A proposed fix, the policy line behind it and a reply ready to go.

Why each part sits where it does

  • AI agent. A refund policy is full of “unless” and “except”.

Reads

  • Refund policy
  • CRM
  • Orders

Writes

  • Ticket queue

The agent

Ticket triage and refundsResolves what is within your limits
Step 3 of 5
Workflow automation: Sends any refund over $250 to the support lead, with the photo, the order and the policy line in one message.
If over $250Person: Approves the $480 refund on order #40219, or declines it with a reason the customer will see.Declined: back to the agent with the reason
What if: 13 hours a month back, worth $6,600 a year

What if you had it

13 hours a month back$6,600 a year

About 10 minutes of gathering on each of 80 larger refunds.

Customize to your business
Starts when
A proposed fix, the policy line behind it and a reply ready to go.
Result
The refund approved or declined, with the reason on the ticket.

Why each part sits where it does

  • Workflow automation. Whether a refund is over $250 is a comparison of two numbers.
  • Person. Above your limit, a refund is money you want someone to sign for.

Reads

  • Ticket queue

Writes

  • Ticket queue
  • Chat
Step 4 of 5
AI agent: Reads the final reply against the account and the policy, and holds it if it offers a credit, date or discount nobody approved.
Workflow automation: Sends the reply, issues the refund on the order and closes the ticket with the reason recorded.Reply sentRefund issuedTicket closed
What if: 30 hours a month back, worth $15,100 a year

What if you had it

30 hours a month back$15,100 a year

About 1 minute checking and closing each of 1,800 tickets.

Customize to your business
Starts when
The refund approved or declined, with the reason on the ticket.
Result
What the customer was told matches what the record says.

Why each part sits where it does

  • AI agent. A promise can hide in a friendly sentence like “we’ll sort that on your next bill”.
  • Workflow automation. Once a reply is cleared, sending, refunding and closing need nobody’s input.

Reads

  • Ticket queue
  • Refund policy

Writes

  • Ticket queue
  • Orders
  • Email

The agent

Promise check on repliesCatches replies that promise too much
Step 5 of 5

Ends with

Answered and closed

The customer has an answer or a refund, and the ticket records why.

Total178h/mo$105k/yr

What changes

Your support team opens each ticket to find the order attached, the problem named and a reply drafted within policy.

  • 3steps automated
  • 4steps for an AI agent
  • 2calls kept with a person

What if you had it

178 hours a month back$104,700 a year in time and savings

For a support team handling about 1,800 tickets a month, at $42 an hour.

Customize to your business

Help articles from tickets

Articles for the questions you keep answering

Starts with

Monday morning

Last week’s closed tickets and survey comments are ready to read.

Workflow automation: At 7am on Monday, pulls every ticket closed last week with its tags, the reply that solved it and any survey comment.
AI agent: Sorts about 400 tickets and comments by the question behind them, so “can’t log in” and “reset link expired” land together.
What if: 13 hours a month back, worth $7,500 a year

What if you had it

13 hours a month back$7,500 a year

About 3 hours every Monday sorting 400 tickets by hand.

Customize to your business
Starts when
Last week’s closed tickets and survey comments are ready to read.
Result
Last week’s questions, grouped and counted.

Why each part sits where it does

  • Workflow automation. It’s a saved report that runs itself every Monday.
  • AI agent. Tags get picked in a hurry. Two tickets tagged differently are often one question, and two tagged alike often aren’t.

Reads

  • Ticket queue
  • Survey responses

Writes

  • Question log

The agent

Feedback themesGroups survey and ticket feedback into themes
Step 1 of 5
If asked 10+ timesAI agent: Checks each repeat question against the help center and flags the ones with no article, or one that shows an old screen.
What if: 9 hours a month back, worth $5,200 a year

What if you had it

9 hours a month back$5,200 a year

About 2 hours a week searching the help center by hand.

Customize to your business
Starts when
Last week’s questions, grouped and counted.
Result
The questions worth an article, ranked by how often they come up.

Why each part sits where it does

  • AI agent. Search finds articles that share the words, which isn’t the same as answering the question.

Reads

  • Question log
  • Help center

Writes

  • Question log

The agent

Help articles from ticketsDrafts help articles from repeat questions
Step 2 of 5
AI agent: Writes the article from the replies that closed those tickets fastest, in your help center’s headings and tone.
Workflow automation: Saves the draft with links to its source tickets and assigns it to the support manager.
What if: 40 hours a month back, worth $23,000 a year

What if you had it

40 hours a month back$23,000 a year

About 2.5 hours saved on each of 16 articles a month.

Customize to your business
Starts when
The questions worth an article, ranked by how often they come up.
Result
A draft article with its source tickets linked.

Why each part sits where it does

  • AI agent. Your team has already written this answer ten times, each a little different. The draft starts from the best of them.
  • Workflow automation. Every draft goes to the same reviewer.

Reads

  • Ticket queue
  • Help center

Writes

  • Shared drive

The agent

Help articles from ticketsDrafts help articles from repeat questions
Step 3 of 5
Person: Fixes any step the product has changed since those tickets, like a renamed settings page, then approves it.Declined: back to the draft with notes
What if: 8 hours a month back, worth $4,600 a year

What if you had it

8 hours a month back$4,600 a year

About 30 minutes of fact-finding saved on each of 16 articles.

Customize to your business
Starts when
A draft article with its source tickets linked.
Result
An approved article, or a draft with notes.

Why each part sits where it does

  • Person. Every customer will read it, and the manager knows what has changed in the product since those tickets were answered.

Reads

  • Shared drive

Writes

  • Shared drive
Step 4 of 5
Workflow automation: Publishes the article, adds it to the reply templates and links it on the tickets it came from.Help centerReply templatesSource tickets
What if: 50 hours a month back, worth $28,800 a year

What if you had it

50 hours a month back$28,800 a year

About 250 fewer repeat tickets a month, at 12 minutes each.

Customize to your business
Starts when
An approved article, or a draft with notes.
Result
The next customer finds the answer before writing in.

Why each part sits where it does

  • Workflow automation. Publishing, adding a template and linking the tickets are three fixed steps.

Reads

  • Shared drive

Writes

  • Help center
  • Ticket queue
Step 5 of 5

Ends with

Article published

An approved article in the help center and the reply templates, before the next customer asks.

Total120h/mo$69k/yr

What changes

The questions customers keep asking become help articles every week, drafted from the replies that already solved them.

  • 3steps automated
  • 3steps for an AI agent
  • 1call kept with a person

What if you had it

120 hours a month back$69,100 a year in time and savings

For a support team closing about 1,800 tickets a month, around 400 a week, at $48 an hour.

Customize to your business

Churn and renewals

Catch quiet accounts early and prep every renewal

Starts with

Every night

Usage, open tickets and renewal dates for every account are pulled together overnight.

Workflow automation: Writes each account’s logins, seats in use, open tickets and renewal date to its CRM record every night.
What if: 33 hours a month back, worth $24,600 a year

What if you had it

33 hours a month back$24,600 a year

About 5 minutes a month pulling numbers on each of 400 accounts.

Customize to your business
Starts when
Usage, open tickets and renewal dates for every account are pulled together overnight.
Result
One current record per account.

Why each part sits where it does

  • Workflow automation. Logins, seats, tickets and dates are numbers a system already keeps.

Reads

  • Product usage
  • Ticket queue
  • CRM

Writes

  • CRM
Step 1 of 5
AI agent: Looks at the last 30 days for logins sliding, the main contact going silent or tickets turning sharper, and scores the risk.
If scored at riskWorkflow automation: Posts the account to its owner in chat, with the reasons and the renewal date attached.
What if: 40 hours a month back, worth $29,800 a year

What if you had it

40 hours a month back$29,800 a year

About 6 minutes a month checking each of 400 accounts.

Customize to your business
Starts when
One current record per account.
Result
A short list of accounts worth a call, each with its reasons.

Why each part sits where it does

  • AI agent. A 60% drop over the holidays means little. The same drop after your main contact leaves means a lot.
  • Workflow automation. Whoever owns the account in the CRM gets the alert.

Reads

  • CRM
  • Ticket queue
  • Email

Writes

  • CRM
  • Chat

The agent

Churn risk alertsSpots quiet accounts before they leave
Step 2 of 5
AI agent: Proposes a check-in, a training session or an offer, depending on why usage fell, and drafts the email that opens it.
Person: Usage is down 60% and the renewal is in 45 days. Books a call this week, or marks it a false alarm.Declined: back to watching
What if: 9 hours a month back, worth $36,700 a year

What if you had it

9 hours a month back$36,700 a year

About 45 minutes of prep on each of 12 save calls a month, plus $30,000 in a renewal a year kept from lapsing.

Customize to your business
Starts when
A short list of accounts worth a call, each with its reasons.
Result
A call booked this week, or the account back on watch.

Why each part sits where it does

  • AI agent. The right move depends on why they went quiet, and the reasons are already on the alert.
  • Person. They know the customer, and an offer sent at the wrong moment can do more harm than good.

Reads

  • CRM

Writes

  • CRM
  • Calendars

The agent

Churn risk alertsSpots quiet accounts before they leave
Step 3 of 5
If renewal in 60 daysWorkflow automation: Opens a renewal task for the account manager 60 days before the contract end date.
AI agent: Puts the year on one page: usage trend, the tickets and how they ended, what they pay, the terms and anything promised.
What if: 47 hours a month back, worth $35,000 a year

What if you had it

47 hours a month back$35,000 a year

About 80 minutes saved on each of 35 renewal briefs.

Customize to your business
Starts when
A call booked this week, or the account back on watch.
Result
A one-page brief ready before the renewal call.

Why each part sits where it does

  • Workflow automation. A date in the contract is a clean trigger.
  • AI agent. The facts sit in a year of tickets, a usage report and a 14-page contract. Someone used to read all three.

Reads

  • CRM
  • Ticket queue
  • Contracts
  • Product usage

Writes

  • CRM

The agent

Renewal prepBuilds each renewal brief ahead of the date
Step 4 of 5
If not yet renewedWorkflow automation: Both go out 30 days before the contract ends.Customer reminderRenewal call
What if: 12 hours a month back, worth $8,900 a year

What if you had it

12 hours a month back$8,900 a year

About 20 minutes of chasing on each of 35 renewals.

Customize to your business
Starts when
A one-page brief ready before the renewal call.
Result
Every renewal gets a conversation before its date.

Why each part sits where it does

  • Workflow automation. Both are set by the contract date, which doesn’t move.

Reads

  • CRM
  • Contracts

Writes

  • Email
  • Calendars
Step 5 of 5

Ends with

Renewed or saved

At-risk accounts get a call in time, and every renewal call starts with a brief.

Total141h/mo$135k/yr

What changes

An account going quiet reaches its manager weeks before renewal, while a phone call can still change how it ends.

  • 4steps automated
  • 3steps for an AI agent
  • 1call kept with a person

What if you had it

141 hours a month back$135,000 a year in time and savings

For a company with about 400 customer accounts and 35 renewals a month, at $62 an hour.

Customize to your business

Compliance AI for Compliance →

Audit request list

Working through the auditor’s request list

Starts with

Request list arrives

The auditor posts 36 requests to the audit portal, each with its own due date.

Workflow automation: Copies all 36 requests from the audit portal into the control register, one row each, with the auditor’s due date.
AI agent: Turns “Q2 user access listing” into control AC-04, April to June, and a system export with a reviewer’s sign-off.
What if: 8 hours a month back, worth $6,700 a year

What if you had it

8 hours a month back$6,700 a year

About 15 minutes to log and match each of 33 requests a month.

Customize to your business
Starts when
The auditor posts 36 requests to the audit portal, each with its own due date.
Result
Every request tied to a control and a period.

Why each part sits where it does

  • Workflow automation. Moving 36 rows from the portal to the register is pure copying.
  • AI agent. Auditors write requests in their own words, and those rarely match the names in your register.

Reads

  • Audit portal
  • Control register

Writes

  • Control register

The agent

Auditor requestsAnswers request lists from the record
Step 1 of 5
AI agent: Searches the evidence folder, accounting and the system logs for the document that answers each request, for the period the auditor named.
Workflow automation: Attaches every file it found to that request’s row, so the question and its answer sit on one line.
What if: 44 hours a month back, worth $37,000 a year

What if you had it

44 hours a month back$37,000 a year

About 80 minutes of searching for each of 33 requests a month.

Customize to your business
Starts when
Every request tied to a control and a period.
Result
A file against most requests, and a short list of the ones without.

Why each part sits where it does

  • AI agent. The file that answers a request could be named anything and saved in any of three places.
  • Workflow automation. Once the right file is found, attaching it is filing.

Reads

  • Evidence folder
  • Accounting
  • System logs

Writes

  • Control register

The agent

Auditor requestsAnswers request lists from the record
Step 2 of 5
AI agent: Opens each file and checks the dates cover the period, it’s the final version, and the sign-off the control needs is on it.
What if: 11 hours a month back, worth $9,200 a year

What if you had it

11 hours a month back$9,200 a year

About 20 minutes opening and checking each of 33 files.

Customize to your business
Starts when
A file against most requests, and a short list of the ones without.
Result
Only final, signed files go forward.

Why each part sits where it does

  • AI agent. A draft and a signed final look the same in a folder. You only tell them apart by opening them.

Reads

  • Evidence folder
  • Control register

Writes

  • Control register

The agent

Control evidenceCollects evidence for every control
Step 3 of 5
If evidence is missingPerson: Agrees who produces each missing piece, like the payroll access review, and by when, so the auditor hears a real date.
Workflow automation: Sends the control owner the request and the date, then a reminder every two days until the file lands.
What if: 12 hours a month back, worth $10,100 a year

What if you had it

12 hours a month back$10,100 a year

About 6 missing files a month, 2 hours of sorting and chasing each.

Customize to your business
Starts when
Only final, signed files go forward.
Result
Each missing file has a name and a date against it.

Why each part sits where it does

  • Person. Putting work on someone’s week while the auditor waits is a decision a person should make.
  • Workflow automation. The owner’s name is on the register, and the reminders run off the calendar.

Reads

  • Control register
  • Org chart

Writes

  • Email
  • Control register
Step 4 of 5
AI agent: Puts each answer together in the auditor’s order: the file, plus a line on what it shows and which system it came from.
If it touches a findingPerson: Reads the three answers that touch open findings before they go, and rewrites the note where the auditor needs more context.Declined: back to the agent with notes
Workflow automation: Uploads each approved answer to the audit portal against its request and marks the row closed in the register.
What if: 14 hours a month back, worth $21,800 a year

What if you had it

14 hours a month back$21,800 a year

About 25 minutes writing up each of 33 answers a month, plus $10,000 in extra audit fees for late answers.

Customize to your business
Starts when
Each missing file has a name and a date against it.
Result
33 answers uploaded at once, and the 3 on findings read by the controller first.

Why each part sits where it does

  • AI agent. A line explaining a file is only right if something actually read the file.
  • Person. An answer on an open finding can change what the auditor writes up. The controller sees those first.
  • Workflow automation. Uploading 36 answers is the same step 36 times.

Reads

  • Control register
  • Evidence folder

Writes

  • Audit portal
  • Control register

The agent

Evidence packs on demandAssembles an audit pack in minutes
Step 5 of 5

Ends with

List closed

Every request answered with a final, signed file, and anything touching a finding read by the controller first.

Total89h/mo$85k/yr

What changes

Control owners are asked once, for one file, by a date. The controller reads only the answers that touch a finding.

  • 4steps automated
  • 4steps for an AI agent
  • 2calls kept with a person

What if you had it

89 hours a month back$84,800 a year in time and savings

For a company facing about 5 audits a year, financial, SOC 2 and customer reviews, some 400 requests in all, at $70 an hour.

Customize to your business

Quarterly access reviews

Who can get into what, checked every quarter

Starts with

Quarter ends

The quarter closes, and every system’s access has to be reviewed and signed off.

Workflow automation: Exports the user list from each system in scope, with every account’s permissions and last sign-in, and matches each one to a person in the HR system.
What if: 20 hours a month back, worth $16,800 a year

What if you had it

20 hours a month back$16,800 a year

About 3 hours exporting and matching each of 20 systems a quarter.

Customize to your business
Starts when
The quarter closes, and every system’s access has to be reviewed and signed off.
Result
One list of every account, in every system, with its last sign-in.

Why each part sits where it does

  • Workflow automation. Every system can export its user list, with permissions and last sign-in.

Reads

  • IT access
  • System logs
  • HR system

Writes

  • Control register
Step 1 of 5
Workflow automation: Marks every account whose owner has left the company, and any that nobody has signed into for 90 days.
AI agent: Compares each person’s access with what others in the same role and team have, and flags the extras, like a sales rep who can still approve payments.
What if: 22 hours a month back, worth $18,500 a year

What if you had it

22 hours a month back$18,500 a year

About a minute on each of 4,000 accounts every quarter.

Customize to your business
Starts when
One list of every account, in every system, with its last sign-in.
Result
A handful of flagged accounts per system, each with its reason.

Why each part sits where it does

  • Workflow automation. Whether someone still works here, or signed in this quarter, is a yes or no in the data.
  • AI agent. Nobody writes down exactly what a role should have. Spotting the odd one out means comparing people who do the same job.

Reads

  • Control register
  • HR system
  • Org chart

Writes

  • Control register

The agent

Access reviewsRuns quarterly access reviews across your systems
Step 2 of 5
Workflow automation: Sends each system owner only the flagged accounts for their system, each with its reason, to keep or remove in one click.
If anything is flaggedPerson: Keeps the access someone still needs, like finance holding a leaver’s login until the handover ends, and removes the rest.Declined: marked for removal
What if: 27 hours a month back, worth $22,700 a year

What if you had it

27 hours a month back$22,700 a year

20 owners each spend about 4 hours less on it a quarter.

Customize to your business
Starts when
A handful of flagged accounts per system, each with its reason.
Result
Owners have ruled on every flag in their own system.

Why each part sits where it does

  • Workflow automation. Who owns each system is already on the register, so each list reaches the right person without anyone sorting it.
  • Person. The owner knows why an odd login exists. The data only shows that it does.

Reads

  • Control register
  • Org chart

Writes

  • Email
  • Control register
Step 3 of 5
Workflow automation: Removes each rejected account where the system allows it, and raises an IT ticket for every one it can’t.
AI agent: Reads the replies on each ticket and follows up the stuck ones, like “shared login, can’t remove yet”, logging them as findings.
What if: 17 hours a month back, worth $32,300 a year

What if you had it

17 hours a month back$32,300 a year

About 20 minutes on each of 150 removals a quarter, plus $18,000 in licenses for removed accounts not renewed.

Customize to your business
Starts when
Owners have ruled on every flag in their own system.
Result
Rejected access gone, and anything that couldn’t be removed logged as a finding.

Why each part sits where it does

  • Workflow automation. The owner has already said no. Removing the account just carries that out.
  • AI agent. Tickets come back with replies in plain words, and “done”, “partly done” and “can’t” need telling apart.

Reads

  • IT tickets
  • IT access

Writes

  • IT access
  • IT tickets
  • Risk register

The agent

Remediation trackingKeeps every finding moving to closed
Step 4 of 5
AI agent: Gathers the user lists, each owner’s answers and proof of every removal into one pack per system, in the form auditors test.
Person: Reads the summary, including any access that couldn’t be removed yet, and signs the review for the quarter.Declined: back to the owners with questions
Workflow automation: Files each pack in the evidence folder, closes the control for the quarter and books the next review.Packs filedControl closedNext review booked
What if: 20 hours a month back, worth $30,800 a year

What if you had it

20 hours a month back$30,800 a year

About 3 hours building each of 20 evidence packs a quarter, plus $14,000 in outside help on reviews not needed.

Customize to your business
Starts when
Rejected access gone, and anything that couldn’t be removed logged as a finding.
Result
The evidence is on file before the auditor asks for it.

Why each part sits where it does

  • AI agent. The proof of who decided what is spread across emails, lists and tickets, in no particular order.
  • Person. The auditor tests the sign-off, and it has to carry a named person.
  • Workflow automation. Filing the packs and booking a date three months out is admin.

Reads

  • Control register
  • IT tickets
  • Email

Writes

  • Evidence folder
  • Control register
  • Calendars

The agent

Control evidenceCollects evidence for every control
Step 5 of 5

Ends with

Reviewed and filed

Each owner has reviewed their system, rejected access is gone and the proof is filed for audit.

Total106h/mo$121k/yr

What changes

System owners decide on the few accounts that look wrong, with a note on why, rather than scrolling a full export.

  • 5steps automated
  • 3steps for an AI agent
  • 2calls kept with a person

What if you had it

106 hours a month back$121,100 a year in time and savings

For a 300-person company reviewing 20 systems every quarter, about 4,000 accounts in all, at $70 an hour.

Customize to your business

Filing calendar

Filings and policy sign-offs, each with an owner

Starts with

First of the month

Each month starts with a look at everything due in the next 90 days.

Workflow automation: Pulls every filing, license renewal and policy sign-off due in the next 90 days from the obligations register.
AI agent: Checks entity records and recent contracts for obligations nobody added, like a sales tax license in a state you just started selling in.
What if: 10 hours a month back, worth $7,800 a year

What if you had it

10 hours a month back$7,800 a year

About 20 minutes on each of 30 new contracts and entity changes a month.

Customize to your business
Starts when
Each month starts with a look at everything due in the next 90 days.
Result
The quarter’s full list, including the one nobody added.

Why each part sits where it does

  • Workflow automation. Everything on the register already has a date, and sorting by date is a query.
  • AI agent. A new obligation shows up as a clause in a contract or a new office address. Nobody adds it as a row.

Reads

  • Obligations register
  • Entity records
  • Contracts

Writes

  • Obligations register

The agent

Filing calendarKeeps every filing on time, with an owner
Step 1 of 5
Workflow automation: Checks every item’s owner against the HR system and flags any whose owner has left or was never named.
If there’s no ownerPerson: Picks who owns the new state license due in 30 days, and anything left behind when someone moved on.
What if: 6 hours a month back, worth $4,700 a year

What if you had it

6 hours a month back$4,700 a year

About 30 minutes checking and reassigning each of 12 filings a month.

Customize to your business
Starts when
The quarter’s full list, including the one nobody added.
Result
Every filing has a person’s name on it.

Why each part sits where it does

  • Workflow automation. The HR system already knows who has left.
  • Person. Giving someone a filing adds to their job, and they should hear it from a person.

Reads

  • Obligations register
  • HR system

Writes

  • Obligations register
Step 2 of 5
AI agent: Fills in the annual report from entity records and last year’s filing, and lists what changed, like a new officer or a moved office.
Workflow automation: Sends the owner the draft, then reminders 30, 14 and 3 days before the due date.
What if: 30 hours a month back, worth $30,900 a year

What if you had it

30 hours a month back$30,900 a year

About 3 hours drafting each of 10 filings a month, plus $7,500 in late filing penalties and fees.

Customize to your business
Starts when
Every filing has a person’s name on it.
Result
A draft with the owner weeks before it’s due.

Why each part sits where it does

  • AI agent. Most filings are last year’s with a detail or two changed, and the change is usually in a board minute or a lease.
  • Workflow automation. Reminder dates count back from the due date.

Reads

  • Entity records
  • Evidence folder

Writes

  • Shared drive
  • Email

The agent

Filing calendarKeeps every filing on time, with an owner
Step 3 of 5
Workflow automation: Sends each person the policies their role signs this quarter through e-sign, and nudges anyone who hasn’t signed after a week.
AI agent: Answers replies like “does this cover contractors?” from the policy itself, and passes anything it can’t settle to the compliance lead.
What if: 33 hours a month back, worth $25,700 a year

What if you had it

33 hours a month back$25,700 a year

About 5 minutes of chasing and questions on each of 400 sign-offs a month.

Customize to your business
Starts when
A draft with the owner weeks before it’s due.
Result
Signatures in, and every question answered.

Why each part sits where it does

  • Workflow automation. Which policies someone signs follows from their role.
  • AI agent. People reply with a question instead of a signature, and the answer is on page 14 of the policy.

Reads

  • Policy library
  • Org chart

Writes

  • E-sign
  • Chat

The agent

Policy attestationsCollects and chases policy sign-offs
Step 4 of 5
Person: Reads the draft, corrects anything wrong and submits it to the state or agency under their own name.Declined: back to the agent with changes
Workflow automation: Saves the filing receipt and signed policies to the evidence folder and sets next year’s due date.Receipt filedRegister closedNext date set
What if: 8 hours a month back, worth $6,200 a year

What if you had it

8 hours a month back$6,200 a year

About 45 minutes checking and filing each of 10 filings a month.

Customize to your business
Starts when
Signatures in, and every question answered.
Result
Proof on file the day it goes in.

Why each part sits where it does

  • Person. The person who signs a filing answers for what it says.
  • Workflow automation. Saving a receipt and moving a date forward a year is housekeeping.

Reads

  • Shared drive

Writes

  • Evidence folder
  • Obligations register
Step 5 of 5

Ends with

Filed on time

Filings are in before their dates, policies are signed and the receipts sit in one folder.

Total87h/mo$75k/yr

What changes

Every filing has a named owner and a draft in their inbox weeks before it’s due, including the ones nobody had on the list.

  • 5steps automated
  • 3steps for an AI agent
  • 2calls kept with a person

What if you had it

87 hours a month back$75,300 a year in time and savings

For a 300-person company with 8 entities in 15 states, about 120 filings a year plus quarterly policy sign-offs, at $65 an hour.

Customize to your business
Your own workflows

Map one of yours in an hour

Get the people who do the work in a room and draw it the same way. It quickly shows where an agent would help, and where it wouldn’t.

  1. 01

    Name the start and the finish.

    What kicks it off, and what done looks like. “Offer signed” to “ready on day one” is a workflow. “HR admin” is too broad to map.

  2. 02

    Put every step in a lane.

    If it moves data or sends a notice, it’s automation. If someone reads something and makes a judgment, it’s a job for an agent. What’s left is a decision.

  3. 03

    Choose the calls that stay human.

    Look for the steps where a wrong answer costs real money or trust, and keep a named person on them. The rest can run on its own.

Our 60-minute workflow audit template walks your team through it. Not sure which workflow to start with? Take the free assessment.

Questions

What people ask us first

What is the AI agent playbook?
A set of everyday workflows from six teams, each drawn step by step and split three ways: what workflow automation does, what an AI agent does, and what a person decides. It is meant to help you spot where agents fit in your own work, and where they don’t.
Do I need an AI agent for every step?
No. In most workflows, most steps are plain automation: moving data, sending the notice, booking the meeting. An agent is worth it on the steps where someone used to read something and make a judgment.
What happens when the agent gets something wrong?
Anything outside the rules you set goes to a person instead of going through, and every workflow here keeps at least one decision with a person. Before an agent runs on its own, it runs alongside your team while people check its work.
Will these work with the tools we already use?
Usually. The workflows use generic systems on purpose, like an HR system, accounting, a CRM or chat, so the same steps apply whichever products you run, as long as they let other software connect.
Can you set one of these up for us?
Yes. Every agent step links to that agent in our catalog. We set it up on your systems and your rules, you try your own version before you commit, and most go live in 3 to 6 weeks.
Next step

Want one of these running for your team?

Tell us which workflow looks like yours. We’ll walk you through the agents behind it, show you how it would work with your systems and rules, and what it would take to get it live in 3 to 6 weeks.