Playbook · for teams and organisations
AI rollout kit
A free, vendor-neutral 90-day plan for bringing AI tools into a team or organisation: what to decide first, who does what, and how to tell whether it worked.
It works with any AI tool. Tick the checklists as you go (the ticks stay in your browser only) and copy them as plain text into your own plan. This is practical guidance, not graded evidence: where studies exist, our rated use cases will say so.
Decide five things first
- Which work comes firstPick two or three tasks that take real time today and where a mistake can be caught. Write down how you do them now and how long they take.
- Who gets which toolsDecide which AI tools are approved, who gets an account, and which kinds of data may go into them. Unapproved tools are where sensitive data leaks.
- The rules of useWrite a short AI use policy: what data is off limits, when a person must check the output, and when to tell others that AI helped.
- What it will costCount licences, usage-based charges, setup and training, and the time people spend checking AI output. That checking time is the cost most often missed.
- How you will know it workedChoose one outcome to measure, take a baseline before you start, and keep a group that hasn't started yet to compare against.
Who does what
| Role | Who | What they do |
|---|---|---|
| Sponsor | A senior leader | Sets the goal, protects time for the pilot and makes the go or no-go call at day 90. |
| Rollout lead | One named person | Runs the plan, keeps the register of AI uses, chairs the weekly check-in and reports results. |
| IT and security | Your IT team or provider | Sets up accounts and single sign-on, data controls and logging; approves tools against your security rules. |
| Privacy, legal and risk | In-house or adviser | Checks the policy, data handling and contracts against the rules that apply to you. |
| Champions | One or two per team | Early users who try tasks first, share what works, and collect problems and questions. |
| Learning and HR | People team | Arranges training and AI literacy for everyone involved, and updates role descriptions where work changes. |
| Everyone using it | All staff in the pilot | Uses the approved tools for the chosen tasks, checks the output and reports errors and near misses. |
In a small organisation one person may hold several roles. What matters is that each job has a name next to it.
The 90-day plan
Days 1–30
Prepare
0 of 8 done
Days 31–60
Pilot
0 of 6 done
Days 61–90
Decide
0 of 6 done
How to tell whether it worked
- One outcomeTime per task, error rate, output volume, a quality score or customer satisfaction. Pick one you already track, so the baseline is real.
- A baselineMeasure that outcome before anyone starts using AI on the task, for long enough to see normal ups and downs.
- A comparisonKeep a similar team or set of tasks working the old way for the pilot, or start teams at different times.
- The full costAdd the time spent checking and correcting AI output to licence and usage costs. A saving that disappears in checking is not a saving.
You measure; we never claim we did. When published studies show whether AI helps with a task, our rated use cases will show how strong that evidence is, so you can compare your own result with it.
Common traps
- No task in mindGiving everyone a tool and hoping for the best makes results impossible to judge.
- No baselineWithout a before measure, any change is a guess.
- Hidden checking timeOutput that must be checked line by line can take longer than doing the work.
- Sensitive data in unapproved toolsStaff use what is easiest; give them an approved option and clear data rules.
- Only measuring enthusiasmSurveys of how people feel about a tool are not the same as measured results.
Official templates and rules
Learn AI for free →Australia: AI policy guide and template, AI register and questions to ask suppliers ↗TemplatesUS NIST: AI Risk Management Framework Playbook ↗GuideUK: AI Playbook for the UK Government (public sector) ↗GuideEU: how organisations build AI literacy (AI Act, Article 4) ↗ExamplesAI laws and rules in your countryRulesEU AI Act: what applies whenRulesFree courses and playbooks for leadersLearn
We link official templates rather than copy them. This page is general guidance, not legal advice.
Written by AI Adoption in our own words, drawing on public guidance from governments and on the rollout material several AI companies publish. Disclosure: AI Adoption uses AI tools in its research and editorial work, including Claude from Anthropic. No company has any say in this page. Spotted an error? Tell us.