Transforming a department with AI takes more than licences and a training session. It means redesigning the department's core workflows so AI is built into how work is done, with clear human checkpoints, shared standards and measurable results. For a typical department, a realistic first transformation cycle takes around six months. Here is what each stage involves.
Stage 1: Map the work (weeks 1 to 2)
List the department's main recurring workflows, typically eight to fifteen. For each, note how often it happens, how long it takes, who does it and what frustrates people about it. Do this with the team, not for them.
You will usually find a few workflows that consume a disproportionate amount of time: reporting, correspondence, documentation, research, scheduling or data handling.
Stage 2: Choose two or three workflows (week 3)
Score each workflow on value (time or quality gain) and ease (how simple it is to change). Pick two or three that score well on both. Resist the urge to transform everything at once.
Stage 3: Set the foundations (weeks 3 to 4)
- Confirm approved AI tools and data rules for the department.
- Take a baseline on the chosen workflows.
- Appoint one or two champions.
- Brief the whole team on what is changing, why, and what it means for them.
Stage 4: Redesign and train (weeks 5 to 10)
For each chosen workflow, design the new version together: where AI does the first pass, where a person reviews, what the quality bar is. Then train the people involved on that specific workflow, using real examples.
Expect the first weeks to be slower, not faster, as people learn. That is normal.
Stage 5: Run, measure and refine (weeks 10 to 20)
Run the new workflows as normal work. Track time, quality and adoption against the baseline. Hold a short review every two weeks to fix problems and share improvements.
Stage 6: Standardise and expand (weeks 20 onwards)
Once a workflow is stable and better than before, document it as the standard process, update onboarding materials and move on to the next workflows on your list.
What department heads often underestimate
- Time for learning. Protect a few hours a week for people to practise.
- Quality checking. AI-assisted work needs clear review steps, especially early on.
- Emotional reactions. Some people will be excited; others will feel threatened. Both need attention.
- Integration needs. Some workflows need AI connected to existing systems, which may require IT support.
Signs the transformation is working
- People describe their work differently, not just "I use AI sometimes".
- New starters learn the AI-supported workflow as the normal way.
- Time saved is being used deliberately, not just absorbed.
- Other departments ask how you did it.
If you are piloting in one department first, see our guide on piloting AI in one department. Leading this kind of departmental change is the core of our Leading AI Innovation & Organisational Transformation course, which includes mapping your own team's tasks and building a 90-day plan.
Frequently asked questions
Which department should be transformed with AI first?
One with a motivated leader, clear recurring workflows and visible pain points. Enthusiastic leadership matters more than which function it is.
How big a team can one manager transform at once?
Most managers can lead change across a team of up to a few dozen people, especially with champions supporting them. Larger departments benefit from phasing by sub-team.
Do we need IT involved?
For approved tools and any integration, yes. For redesigning workflows around general AI assistants, IT support is usually light.