Most people meet AI alone: their files, their inbox, their calendar. It’s genuinely useful, right up to the moment the next step belongs to someone else. Call that single-player. Multiplayer is when the team and its AI work from the same record.
Single-player
Everyone asks their own AI.
Three private chats, three walls, the same question asked three times. Each assistant knows its owner’s corner and nothing crosses the desk.
Multiplayer
Humans + AI refer to the same data.
One person fills a gap, AI completes the next line, a colleague watches it happen. One record, everyone at it together, each inside their own access.
Left, labelled single-player: three separate chat windows divided by dashed walls, each person asking their own AI and getting a question mark back. Right, labelled multiplayer: five people and the AI mark around one shared Northstar record, with cursors filling in a missing line together.
Why it happens
AI arrived one seat at a time.
Most businesses adopted AI the way they adopted laptops: one licence per person. Each assistant knows one person’s slice of the work, and only that slice. That works until a fact lives in someone else’s assistant, two records disagree, or nobody wrote it down at all.
Sales’ AIClosed the deal
Northstar
Start date1 Oct
Monthly fee$2,400
ReportingIncluded, promised in the pitch
Billing contactNot recorded
Ops’ AIOperations and legal
Northstar Group Ltd.
Start date14 Oct
Term12 months, fixed scope
ReportingNot in the signed scope
Billing contactNot recorded
Finance’s AIFinance
NS Group / C-1048
Start date?
Monthly fee$2,000 after discount
Payment termsNet 30
Billing contactNot recorded
?Three names, two start dates, two prices, and nobody knows who receives the invoices.Each assistant is sure of its own slice. None of them can see the disagreement.
One customer record torn into three slices, one for each personal assistant. The sales slice calls the customer Northstar and knows what the buyer is worried about; the operations slice calls it Northstar Group Ltd. and holds the signed terms; the finance slice calls it NS Group and holds the invoice. The start dates, the monthly fee and whether reporting is included all disagree between slices, and the billing contact is recorded nowhere. Below the slices, a line sums it up: three names, two start dates, two prices, and nobody knows who receives the invoices.
Every assistant sees a slice. Nobody’s assistant sees the customer. If your week has a few of those moments, your AI is single-player, whatever the licence says.
The problem
Workflows across teams rely on “ask so-and-so…”
When one team hands work to the next, the context stays behind. Sales lands a new customer, and weeks of calls, notes and AI chats live in one person’s corner: what the buyer is nervous about, why it was scoped this way, which promise closed the deal.
Sales
Closed the deal. Holds the whole story.
Back on their desk???
Week 1
OOps + legal
get the contract
“Northstar Group Ltd.” Is that the same customer as the deal?
Week 2
DDelivery
get a start date
The CRM says 1 Oct, the contract says 14 Oct. Which is real?
Week 3
FFinance
get an invoice to send
Who actually receives this? There’s no billing contact anywhere.
Three handoffs. Three questions. One desk.The salesperson on the left. To their right, three weekly stages: operations and legal get the contract and ask whether Northstar Group Ltd. is the same customer; delivery gets a start date and asks which of two dates is real; finance gets an invoice to send and asks who receives it. Each question travels back along a rail to the salesperson’s desk.
Personal AI makes each person a little faster, and turns the person who knows the most into everyone else’s help desk.
The shift
Same facts. One record. Everyone reads from it.
Nothing moves out of your existing systems. The slices get linked into one record of the thing the work is about, which points back at each system, and everyone, people and AI alike, reads from that one place.
One customer recordNorthstar Group Ltd.Linked: NorthstarNS Group / C-1048
Everyone reads from it
What the buyer is worried aboutReporting for their leadership reviewSales notes
Start date14 OctSigned agreement
The signed termsFixed scope, 12 monthsContract
First invoice$2,400 · Net 30Accounting
Who receives the invoices?Missing · flagged to salesOwner: sales
Margin on the deal••••Finance and leadership only
Same facts, one place. Every answer says where it came from. Gaps show up with an owner instead of hiding.The three slices reassembled into one Northstar Group Ltd. record, with the other two names linked to it. Each fact shows its value and the system it came from: sales notes, the signed agreement, the contract, accounting. The billing contact is marked missing and flagged to sales as its owner. The margin is hidden from everyone but finance and leadership. A row of people and the AI chip read from the same record.
F
Finance asks
Who gets Northstar’s invoices?
The shared AI answers
Nobody yet. The billing contact isn’t on the CRM record or in the agreement. Sales owns that one and it’s been flagged to them. Once they confirm, it lands in accounting on its own.
Source: CRM contact record · signed agreement
When someone asks a question, the answer arrives with a source instead of a shrug.
Inside the workflow
Think of it as an assembly line. AI takes the steps that suit it, not the whole line.
A handoff is work moving from one pair of hands to the next. You choose which steps AI takes, usually the reading and the drafting. People keep every step that needs judgement or a signature.
Step 1SalesConfirm the dealA person
Step 2OpsAIRead the agreementPull scope, terms, contacts
Step 3Ops + legalReview the commitmentsA person approves
Step 4DeliveryAIDraft the kickoff planFrom the approved scope
Step 5DeliveryStart the workA person
Step 6FinanceAIPrepare the invoiceFrom the contract. Finance approves.
Six steps, split by what they need. AI takes the three that are reading and drafting: pulling the scope from the agreement, drafting the kickoff plan, preparing the invoice. People keep the three that need a decision or a signature: confirming the deal, approving the commitments, starting the work.Six numbered steps in a row: sales confirms the deal, the agreement is read, operations and legal review the commitments, the kickoff plan is drafted, delivery starts the work, finance sends the invoice. Every step begins with a person; then AI is placed on steps 2, 4 and 6, the reading and drafting steps, and the work runs along the line.
What changes
Stop chasing each other for updates. Build a system for AI + humans.
Here is what a handoff looks like with one record underneath it.
Fix it once.
Sales confirms the billing contact in the CRM. It lands in accounting and the project record on its own. If two systems disagree, someone sees the difference before it spreads.
Approve it, and it moves.
Operations approves the scope AI pulled from the agreement. That approval is the signal: AI drafts the kickoff plan and delivery is told. Anything missing goes to the one person who can answer it, not to everyone.
The lesson stays.
The team adds a billing-contact check to onboarding and tests it. The next customer’s handoff has it from day one, whoever is on the desk that week.
Over time
People change. The knowledge shouldn’t leave with them.
People go on leave, get promoted, move on. In a single-player setup, the reasons behind their decisions leave with them. In a multiplayer one, those reasons live in the shared record, so the next person starts with the story instead of a blank page.
Single-player
New account lead · first dayNorthstar
Where things left off
?What was promised? Only the previous lead knows.Their private chat. No access
?Who receives the invoices? Not in any record.Ask… someone
?Kickoff date? Two dates, no owner.CRM says 1 Oct, contract says 14 Oct
Everything depends on one person.
Months of context lived in one person’s chats and inbox. When they move on, the next person inherits a signed contract and a list of questions, and takes months to get up to speed.
Multiplayer
New account lead · first dayNorthstar
Where things left off
Reporting was promised for their leadership review.Proposal, sales, March
Billing contact confirmed: accounts team.CRM, confirmed by sales
Still open: kickoff date. Owner: delivery.Project tool
New people are useful from day one.
The record kept the decisions and the reasons behind them. A new hire reads the story instead of reconstructing it, and nobody has to be the only one who knows.
Left, single-player: a new account lead’s first-day view of Northstar with three open questions and nowhere to look, because the context lived in the previous lead’s private chat. Right, multiplayer: the same first day, with AI listing where things left off, each line with its source.
The business keeps what it learned, whoever is on the desk.
Common myths
Four myths about multiplayer AI.
You might think
“It’s a shared drive with AI on top.”
Actually
A shared drive holds files. This holds facts, each with an owner and a source, and the AI works inside the process rather than searching a pile of documents after the fact.
You might think
“More data means more made-up answers.”
Actually
The opposite, when it’s built right. Every answer has to point at a record. When the record is missing, the honest answer is “nobody knows yet, and here’s who owns it,” which beats a confident guess.
You might think
“We’d have to replace our systems.”
Actually
No. The CRM stays the CRM and accounting stays accounting. The shared record points at them, and connectors carry facts between them so people stop retyping.
You might think
“Everyone has to use the same chat tool.”
Actually
No, and most of the value isn’t chat at all. It’s the steps AI runs inside the workflow, which nobody opens a chat window for. People keep the assistants they like.
Keeping it safe
A few rules, set once. Then everyone moves faster.
One record doesn’t mean everyone sees everything. It’s the same good sense you already apply to people, written down for the AI: who sees what, what it may do on its own, and who looks after it.
01
Access follows roles.
Everyone sees what their role allows. The AI plays by the same rules, in what it reads and in what it says.
02
AI drafts, people approve.
The AI prepares the work. A person says yes to what matters. Sources and decisions are logged, so anyone can check.
03
Someone owns it.
A process with a name next to it keeps improving after launch, and gets fixed quickly when it needs to.
04
Knowledge stays yours.
Your records, instructions and improvements belong to your business, whichever AI provider you use next.
How to start
Start with one handoff, not the whole company.
The businesses that get this working don’t begin with a platform decision. They begin with one handoff everyone already complains about.
Day one
Pick the handoff that ends in “ask someone.”
Sales to delivery. Intake to underwriting. Order to invoice. The one where the same question gets asked every time.
Week one
Draw the steps and decide who stands where.
Which steps are reading and drafting? Which are judgement? Where does a person need to say yes before anything moves?
Weeks two and three
Prototype it on your real records.
One team, actual data. If it doesn’t hold up there, it won’t hold up in production. That’s what an AI Jumpstart is for.
Next step
Stop guessing where AI belongs. We’ll show you.
An AI Jumpstart is a short, fixed-scope project. We get inside your operations, map where AI is worth the effort and where it isn’t, and prove it with a working prototype on your real records.