The short answer.
When everyone uses AI their own way, the trouble shows up between teams. Handoffs stop matching because each team changed its output on its own, customers get two answers to the same question because two teams asked AI from two different sources, and checks get skipped because each person assumed someone else checked. The fix is a plan for the work between teams: for each handoff, one agreed format, one source for the facts and one named person who checks.
Experimenting was the right way to start, and each person’s use of AI usually works well inside their own job. Our AI use is all over the place covers the first week of taking stock inside each team.
The trouble shows up between teams, not inside them.
AI used in one person’s own way rarely causes trouble inside that person’s job, because that person knows what they asked for and reads what comes back. The trouble starts when the output leaves the desk. The next team receives work shaped by a prompt it never saw, from a source it does not know.
Inside each team, everything looks fine. Between teams, the pieces stop fitting.
Three complications you will recognise.
- Handoffs that no longer match.A team starts preparing its weekly summary with AI, and the summary changes shape, and the one figure the next team relied on is folded into a paragraph. The receiving team rebuilds what it needs by hand, or misses it.
- Two answers to the same customer question.Sales asks its assistant about delivery terms using an old price list. Customer service asks using the current order system. The customer hears two answers from one company, and both people sent theirs in good faith.
- Checks skipped because each person assumed another checked.Before AI, the person who wrote a document also read it closely. Now one person prepares a draft with AI and passes it on, assuming the next person reviews it, and the next person assumes it was reviewed already. The work goes out with nobody having read it closely.
All three complications come from the same gap: the work between teams has no owner. How do we keep quality up when AI prepares the work? sets out how a named check is built into a workflow.
What it costs the company.
The cost of undirected AI use between teams is paid in rework, in customer trust and in the attention of your best people. None of that cost shows on the AI bill.
- Rework.The receiving team redoes work the sending team believes is finished.
- Customer trust.A customer who hears two answers starts checking everything you send.
- Your leads’ attention.Team leads spend their week tracing who said what.
- Your sight of the business.You no longer know how a quote, an order or a reply is put together, so you cannot say where to fix it.
What happens when a company uses AI without a plan? sets out the cost inside each role: repeat work that never moves onto AI. Between teams, work moves onto AI unevenly, so the parts stop fitting.
How it feels from the inside.
From the inside, undirected AI use between teams feels like a slow loss of footing. Your people feel it first, and you feel it last.
It starts with excitement. People find that an assistant drafts a reply well, and they are right to be pleased. Then comes friction. Summaries look different every week. A customer calls back to say they were told something else. Someone says “I thought you checked that,” and the other person says the same.
Then comes a quieter doubt. People stop trusting the work that arrives from the next team and check it again themselves, so the effort AI saved in one team is spent in the next. Good people feel blamed for mistakes that nobody owned.
For you, the experience is confusing in a particular way. Every team tells you AI is helping, and the business feels harder to run. Both are true. The confusion is the sign that the work between your teams needs a plan, and it is not a sign that anyone has failed.
A plan settles the work between teams.
A plan settles the work between teams by writing down, for each handoff, what passes from one team to the next, in what format, from which source, and who checks it before it moves on. Each team keeps its own way of working inside its own job.
FullyPowered Planner is where the plan lives. You set out your teams, their roles and the work between them, choose what gets built first and name who owns each piece, and your colleagues add their own teams.
From the plan, the handoffs that matter most become workflows: run the same way by everyone, in your own ChatGPT, Claude, Copilot or Gemini. Where the method sits in one person’s head, FullyPowered Studio captures it as a workflow. A person decides; the workflow prepares. How to manage AI across teams without everyone doing their own thing sets out who holds the plan and what stays with each team.
A handoff sheet to copy.
Copy these headings onto one page and fill in one block for each handoff between two teams. Write “Unknown” where nobody knows.
- From and to.The sending team and the receiving team, named for the work.
- What passes.The piece of work that changes hands: a quote, an order, a weekly summary, a reply.
- The agreed format.The headings or fields the receiving team needs, in the order it needs them.
- The source of the facts.The one system or file the facts come from, such as your order system or the current price list.
- Who checks.One name on the sending team, who reads the output before it leaves.
- When it does not match.What the receiving team does with work that arrives wrong, and to whom it goes back.
- Owner of the handoff.One person, by name, who keeps the format and the source current.
The rule for the sheet: a handoff with no name in row five or row seven goes first on the plan.
One manufacturer maps three handoffs.
Take a 150-person manufacturer with sales, customer service and purchasing teams. The operations director filled in the handoff sheet with the three team leads.
Illustrative example.
- Purchasing to customer service.Purchasing sends a weekly summary of supplier delays. One buyer began preparing it with AI and the layout changed; customer service lost the part numbers it answers from. The decision: one agreed format, part numbers first, owned by the purchasing manager.
- Sales and customer service, one customer question.Customers asked both teams about delivery dates. Sales drafted answers from the lead times in its quoting spreadsheet; customer service drafted them from the order system. The decision: the order system is the one source for delivery dates, and the customer service lead owns the delivery answer for both teams.
- Sales to purchasing.Sales passes on confirmed orders that need special materials. A sales rep drafted the notes with AI and assumed purchasing would check the specification; purchasing assumed sales had confirmed it. The decision: the sales lead reads every special-materials order note before it leaves sales, and purchasing sends back any note without that sign-off.
The delivery answer became the first workflow on the plan, because it touched two teams and every customer.
What is different in a company of under 200 people.
In a company of under 200 people, the work between teams runs through people who know each other, so the complications stay hidden longer and settle quickly once named. Most published advice is written for large enterprises, where each handoff already has a process owner.
- Handoffs are informal.A question across the corridor replaces a form, and nobody notices when AI changes what the corridor covered.
- One person is often the handoff.A single coordinator may carry every order from sales to purchasing. Treat that role as a team on the sheet.
- The owners fit around one table.The leads who own your worst handoffs can fill in the sheet together, in the same room.
Start with the handoff that causes the most phone calls this month. How organizations plan their use of AI covers the full method: plan it, develop the workflows, deploy them. Set out your teams and the work between them in FullyPowered Planner, and when you want help with the first workflows between your teams, we develop them with you in a 3-Day Workflow Sprint.