AI Team Planning Worksheet

How to use it. Work through the four parts with the people who will own the result: usually the executive accountable for AI, whoever runs product, and HR. Part 1 takes stock of who holds each role today. Part 2 names an owner for each piece of agent work, one agent at a time. Part 3 sets your hiring order. Part 4 maps the route from each person's current role to the next one. Titles vary from firm to firm; fill in the work, not the title.

Part 1. Role inventory

For each role, write who does this work today (a name, a part of someone's job, a vendor, or nobody), and what changes as agents take on more of the work.

Role	Who holds it today	Share of a job	Gap or next step
Roles that still matter
Accountable AI executiveOwns the AI portfolio and the decision to stop an agent
AI product ownerDecides what an agent is for, what it may not do, and when it hands back to a person
Data engineerPipelines, access rules and quality checks the agents act on
Data scientistLess routine analysis; more tests of whether agent output can be trusted
ML engineerFrom shipping models to running agent systems
Governance specialistMaps systems to the rules that apply, runs reviews, keeps the records
TranslatorKnows enough of the technology, the business and the people to get a system adopted
Roles that are new
Agent orchestratorDesigns the workflow: which agent does what, with which tools, within which limits
Agent evaluatorBuilds the tests and sample checks that decide whether an agent keeps running
Agent reliability engineerMonitoring, drift, limits and the switch that stops an agent
Oversight designerDecides where a person approves, where a person reviews after the fact, and where the agent acts alone

Part 2. Who owns what, for one agent

Agent name:

Write a name in each box. R is the person who does the work; A is the one person who answers for it. Each row gets exactly one A. At a small company one person may hold several columns. That’s fine as long as every row has an owner.

Agent work	Responsible (R)	Accountable (A)	Consulted	Informed
Design the workflow
Set its authority limits
Evaluate its outputs
Run and monitor it
Stop it
Answer for the outcome

Check: can the person in the A box for "Stop it" have the agent switched off within a day, without asking anyone else? Yes No

Part 3. Hiring order for a company of about fifty

Tick each step when it is done. Move to the next only when the one before has paid for itself.

Tools for everyone. Give every person the AI assistants that fit their work, with a short written policy on what data may go into them. No hire needed.
A scoped review of where AI pays. Rank three to five use cases by value and by how clean the data behind them is. Name the one you will fund first.
First full-time hire: an AI product owner. Someone who can turn a business problem into a scoped use case and decide what an agent may and may not do.
Second hire: a data engineer, if the use case you funded is held back by data quality or access.
Third hire: a governance specialist if you work in a regulated sector, otherwise a translator who gets the work adopted across the company.
Only then: a data scientist or ML engineer, when a use case needs a model you cannot buy.

Part 4. The route from today's role to the next

For each person on the team, write the role they hold, the role it is moving toward, and what they need to get there. People stay when they can see the route.

Person	Role today	Role next	What they need (training, a project, a mentor)

Common routes: data scientist to agent evaluator; ML engineer to agent reliability engineer; product manager to oversight designer.

Three questions to ask each quarter:

How much of each technical person's week goes on cleaning or finding data?
What has each of them built that real users now use?
Does a senior business owner know what each of them does, and say so?