Rolling out AI works when you start from people's real work, not from the tools. In 30 days you can pick the right assistant, write a one-page usage policy, train each team on three to five tasks they do every week, and measure adoption and time saved. Week one is discovery, week two is setup and policy, week three is training, week four is measurement and next steps.
Why most AI rollouts fail
Companies buy licenses, send a launch email and wait. A few enthusiasts use the tool. Everyone else tries it once, gets a generic answer and goes back to their old way of working.
The problem is not the tool. It is that nobody connected it to the specific tasks each person does every week. The plan below fixes that.
Week 1: Discover where AI saves time
Interview two or three people per team for 30 minutes. Ask:
- What tasks do you repeat every week?
- What do you postpone because it is tedious?
- Where do you write, summarize, search or reformat information?
Collect the answers in a simple list with the task, the team, how often it happens and how long it takes. Then rank by hours per month. You will usually find 10 to 20 good candidates such as proposals, meeting notes, reports, support replies, job descriptions and research.
Week 2: Choose tools and write the policy
Pick the assistant
Test two assistants on five real tasks from your list. Compare output quality, how well they fit your existing software, admin controls and price per seat. Business plans of ChatGPT, Claude, Gemini and Copilot all offer admin controls and do not train on your data by default.
Write a one-page AI policy
Keep it short enough that people actually read it:
- Approved tools. Which assistants and plans are allowed.
- Data rules. What can go in (public info, internal docs) and what cannot (for example health data or client data without agreement).
- Check before use. AI output is a draft. A person is responsible for what is sent or published.
- Disclosure. When to mention that AI was used, for example in client deliverables.
- Questions. Who to ask when unsure.
EU AI Act
The EU AI Act expects organizations to ensure sufficient AI literacy among staff who use AI. A clear policy plus role-based training is the simplest way to show you take this seriously.
Set up shared assistants
Create shared custom GPTs, Claude Projects or Gems for your most common tasks, loaded with your tone of voice, templates and reference documents. People get good results from day one without writing long prompts.
Week 3: Train by role, on real tasks
Skip the generic "intro to AI" lecture. Run 90 to 120 minute workshops per team, where everyone works on their own tasks from the week 1 list:
- Sales: proposals from call notes, account research, follow-up emails.
- Marketing: briefs, content drafts, repurposing, ad copy variations.
- Operations: summaries of documents and contracts, process docs, spreadsheets.
- Support: reply drafts from the help center, ticket summaries.
- Management: reports, meeting prep, decision memos.
Teach three skills that apply everywhere: give context, show an example of good output, and ask the assistant to check its own work against the source.
Week 4: Measure and plan next steps
Measure three things:
- Adoption. Weekly active users per team, from the admin panel.
- Time saved. Ask people to log minutes saved on their top three tasks.
- Quality. Collect examples where AI helped and where it failed.
Share the results with the whole company, including the honest failures. Then pick the two or three tasks that came up most and turn them into automations, so they happen without anyone opening a chat window. That is where AI automation takes over.
A simple scorecard
| Metric | Week 1 baseline | Day 30 target |
|---|---|---|
| Weekly active AI users | Measure | Most of the team |
| Tasks with a shared assistant | 0 | 5 to 10 |
| Policy read and accepted | 0% | 100% |
| Logged hours saved per week | 0 | Track and report |
Want help running this plan?
We run AI adoption programs for teams in English, Dutch, French and Arabic: discovery, tool setup, policy, training and follow-up. See our AI adoption and training service or book a free call.