Skip to content
Business Operations agent

Internal questions

Ask in Slack or Teams how to raise a PO, book travel or order a laptop, and the answer comes from your own docs with a link to the page. When two docs disagree, a person settles it.

  • $12,500 fixed
  • Live in 3 to 6 weeks
  • Yours to keep
The review screen for Internal questions: four items, three done, and one (Docs disagree) waiting for the ops manager, who is asked: The travel policy says book 14 days ahead and the wiki says 7. Answer with the policy?
How it works

Quick questions get quick answers, straight from your own docs.

Here’s a typical run, step by step: what automation handles, what the agent does, and the calls that stay with your team.

The workflow for Internal questions, in 5 steps. It starts with “Someone asks in chat” and ends with “An answer with its source”. Each step lists what automation, the agent and your team do there.
Internal questionsA typical run
Starts withSomeone asks in chatA consultant on the delivery team asks in the help channel how far ahead to book flights for a client trip.
The agent: Reads “how far ahead do I book flights?” as a question for the travel policy, and notes that it’s for a client trip.
~10h/mo

Step 1 of 5Read the question

  • The agent

    Reads “how far ahead do I book flights?” as a question for the travel policy, and notes that it’s for a client trip.

    People ask the same thing a dozen ways, and few of them use the words in the policy’s title.

Reads
  • Team chat
Writes
  • Question log

~10 hours a month back

The agent: Searches your policies and wiki and finds two pages: the travel policy says 14 days ahead, and the ops wiki says 7.
Automation: Looks up who owns each page and when it last changed: the policy in May, the wiki page in 2023.
~18h/mo

Step 2 of 5Find the answer

  • The agent

    Searches your policies and wiki and finds two pages: the travel policy says 14 days ahead, and the ops wiki says 7.

    The answer is usually in a paragraph, not a title, and sometimes it’s in two places.

  • Automation

    Looks up who owns each page and when it last changed: the policy in May, the wiki page in 2023.

    The owner and the last-changed date are stored with every page.

Reads
  • Wiki
  • Shared drive
Writes
  • Question log

~18 hours a month back

Where teams differStarting with a few spaces, like travel and purchasing, is common. Wider reading can come later.

Automation: When two pages give different answers, it holds the reply and asks the ops manager, with both pages side by side.
IfYour team: Confirms the policy’s 14 days is right and asks the wiki page’s owner to fix it, or says the wiki is the current one.
~3h/mo

Step 3 of 5Settle the conflict

  • Automation

    When two pages give different answers, it holds the reply and asks the ops manager, with both pages side by side.

    Two different answers on file is a fact it can see, not something to judge.

  • Your teamIf two docs disagree

    Confirms the policy’s 14 days is right and asks the wiki page’s owner to fix it, or says the wiki is the current one.

    Choosing which rule stands is a policy call, and it belongs to a person who can make it.

Reads
  • Question log
Writes
  • Question log

~3 hours a month back

The agent: Replies in the thread: book at least 14 days ahead in the travel tool, with the steps and a link to the policy page.
~20h/mo

Step 4 of 5Answer in the thread

  • The agent

    Replies in the thread: book at least 14 days ahead in the travel tool, with the steps and a link to the policy page.

    A good answer is two lines in plain words, not the whole policy pasted in.

Reads
  • Wiki
Writes
  • Team chat

~20 hours a month back

Automation: Logs the question and its answer, and adds anything it couldn’t answer to a weekly list for the ops team to write up.
~4h/mo

Step 5 of 5Log the gaps

  • Automation

    Logs the question and its answer, and adds anything it couldn’t answer to a weekly list for the ops team to write up.

    Counting which questions repeat, and which went unanswered, is tallying.

Reads
  • Question log
Writes
  • Question log
  • Email

~4 hours a month back

Where teams differBusy teams read the gap list weekly. Quieter ones look only when a topic keeps coming back.

Ends withAn answer with its sourceA reply in the thread with the steps and a link to the page, logged so you can see what gets asked.
Total~55h/mo

What changes

Routine questions in chat get an answer from your own docs, with a link. Only the real gaps, and docs that disagree, reach a person.

  • 3steps automated
  • 3steps for the agent
  • 1call kept with your team

Pick any step to see what happens there and why it sits in that lane.

What it could be worth

~55 hours a month back

$31,700 a year in time

For a 300-person company asking about 400 how-do-I questions a month, at $48 an hour. We work out yours from your own volumes before you sign.

Let’s build your agent
What we customize with you

One proven agent, plugged into how you work.

Every team does this job a little differently, so we set these five with the people who do it today. We’ve done this in dozens of businesses, so we know where teams usually differ and ask about those first.

The agent

Same for every client
  • Follows the same proven steps
  • A person on every exception
  • Every decision logged
  • Tested on your past work first
  • Never answers from outside your own docs
  1. Your systems

    Where the work comes in, and where the result goes.

    For this agentQuestions from Slack, Teams or a help inbox, answers from your wiki, shared drive and policies.

  2. Your rules

    The limits and exceptions it works to.

    For this agentWhich docs it reads, which topics it answers, and what always goes to a person.

  3. Your team

    Who signs off, and who covers when they’re away.

    For this agentThe ops manager for gaps and for docs that disagree, each page’s owner for the fix.

  4. Your channels

    Where your team hears from it.

    For this agentReplies in the thread it was asked in, and a weekly list of topics it couldn’t answer.

  5. Your output

    What it hands back, in the format you already use.

    For this agentAn answer with a link to its page, or the question handed to the right person.

It answers from what’s written down. HR questions are the Policy and leave questions agent, requests that need something done go to Request intake, and fixing a doc stays with its owner.

What you get

The agent, and everything that keeps it working.

  • The agentRunning on your systems, under your rules.
  • The review screenWhere the exceptions land for your ops team and the owner of each doc.
  • Your rulebookYour limits, exceptions and who signs off, in one place, often for the first time.
  • A test reportYour past internal questions, run through it and checked against the number.
  • The runbookHow it works, what to check, and what to do if it stops.
  • 30 days of tuningAfter it goes live, through a full month of your real internal questions.
The number we agree

Questions a week your ops team answers by hand. We measure it before we start, and again at day 30.

How we get there

Live in 3 to 6 weeks. About seven hours of your time.

We do the building. Your side: access, your wiki, your policies and three months of questions from your help channel, and a short weekly check-in.

  1. Before kickoff

    We send the access checklist.

    You, 2 hours. Access, an owner, and your wiki, your policies and three months of questions from your help channel.

  2. Weeks 1 to 2

    We connect your systems and set it up.

    You, 1 hour. Walk us through your rules.

  3. Weeks 2 to 4

    We test it on your past internal questions.

    You, 1 hour. Review what it flagged.

  4. Weeks 3 to 6

    It runs beside your team, then goes live.

    You, 1¼ hours. Work from the review screen.

Then 30 days of tuning, with a short check-in each week, about two hours of yours. At day 30 we show you the number and hand everything over.

After it’s live

It keeps running whether or not we stay.

  • It runs in your accounts

    Your cloud, your systems, and an AI account in your name, billed by the provider at their rates.

  • Changes in the first 30 days are included

    A new rule, a changed limit, someone new signing off. That’s what tuning is for.

  • After that, your call

    Your team edits the rulebook, the care plan covers it, or a bigger change is quoted first.

Questions

About this agent.

What if the answer isn’t in our docs?
It says it doesn’t know, hands the question to your ops team, and adds the topic to a weekly list so someone can write the page.
Can it see documents people shouldn’t?
It only reads the folders and spaces you choose, and it can be set to answer each person only from pages they can already open.
Will it edit our wiki?
No. It can draft a fix for a page that’s out of date, and the page’s owner makes the change.
Our process isn’t quite like this. Will it still work?
That’s the usual case. The job looks the same in most businesses, but the details never do. We map how your team does it today, then set the rules, systems and approvals to match. That’s what most of the 3 to 6 weeks are for.
How do we agree the number?
Together, before any work starts. For this agent it’s the questions a week your ops team answers by hand. We measure it before we start, and again at day 30.
What does it cost to run once it’s live?
Hosting and AI, paid straight to the providers, not through us. It usually runs 5x to 25x cheaper than doing the same work by hand, and we work out your figure from your real volumes before you sign.

Where do people ask their how-do-I questions today?

Tell us how your team handles internal questions today. We’ll show you this agent on examples like yours, and whether it fits the fixed price.

New to Switchboard Agents? How it works, what’s included, and what it costs.
  • 3 to 6 weeksto live, then 30 days of tuning
  • $12,500fixed, per agent
  • Yours to keepno license fees
See the service
Switchboard Agents$12,500 fixed · live in 3 to 6 weeks How it works Let’s build your agent