What a 90-day AI capability rollout looks like (week-by-week)
Capability is not a single event. A workshop changes a day, a training video changes an hour, but neither changes how a team works. Capability is built through structured repetition over weeks. The kind of practice that turns 'I tried that thing' into 'this is how we work now.' Ninety days is the shortest window in which we have seen capability take root across a team of 20 to 50. This is what a 90-day rollout looks like week by week, what to measure at each phase, and the three things that consistently derail it.
Why 90 days
The 90-day window is not arbitrary. It is the shortest period in which a team can move through three distinct stages: learning the tools (weeks 1-4), applying them to real work (weeks 5-8), and embedding new ways of working as the default (weeks 9-12). Shorter than 90 days and you skip embedding. The team learns and applies, then drifts back to old workflows within months. Longer than 90 days and momentum decays. The program becomes background noise. Ninety days is also short enough for executive sponsors to commit to without renegotiating priorities, and long enough for measurable behaviour change.
Weeks 1-4: Foundations and tool standardisation
The first four weeks are about standardisation, not innovation. Three things happen.
Week 1 is the kickoff workshop. The whole team in the room (or on the call) for a day. The kickoff covers the tool landscape, the five prompt patterns, and three workflow rebuilds per participant. It sets the shared vocabulary the rest of the 90 days depends on. Without a strong kickoff, the rest of the program is fighting confusion.
Weeks 2-4 are tool standardisation and practice. The team picks one writing tool, one research tool, and one meeting tool as the team standard. Everyone runs at least one prompt patterns exercise per week: small, low-stakes practice. Champions are named in each functional area, typically one champion per five to eight people. Champions get a 30-minute coaching call each week to surface what is working and what is not.
Weeks 5-8: Workflow rebuilds
Weeks 5 to 8 are the operational core of the program. Each participant identifies one high-volume workflow they own, then rebuilds it with the team's standard tools and the five prompt patterns. The rebuild is not a one-shot project. It moves through four checkpoints.
Week 5: identify the workflow, measure baseline time and quality. Week 6: build the prompt and tool combination. First version of the rebuilt workflow ships internally. Week 7: refine based on usage. Champions review what is working across their function. Week 8: ship the rebuilt workflow as the team's new default. Measure new time and quality against baseline.
The output of this phase is concrete. Each participant has shipped one rebuilt workflow with a measured before-and-after. Across a team of 30, that is 30 measurable wins. That is what makes the embed phase possible.
Weeks 9-12: Embed
Weeks 9 to 12 are about consolidating the new ways of working before momentum decays. Three things happen.
Champions run weekly 30-minute team sessions sharing wins, prompts, and patterns. These replace the live coaching calls and create the internal community that keeps capability alive after the program ends.
The team prompt library, built incrementally through weeks 1-8, gets formalised. Naming conventions, an owner, a review cadence. The library has to become a thing people maintain or it dies in a shared folder.
Each participant commits to a single 'practice workflow' for the 30 days after the program: something they will do every week using the team's standard tools and patterns. Practice maintenance is the difference between capability and a brief uplift.
The three KPIs that matter
Most AI training programs measure the wrong things. Satisfaction scores and 'felt more confident' are vanity. The three KPIs that matter are concrete.
Workflows rebuilt: count of distinct workflows that have moved from baseline to a new AI-assisted version with measured time savings. Target: one per participant by week 8.
Adoption rate: percentage of the team using the standard tools at least three times per week, measured at weeks 4, 8, and 12. Target: 80% by week 12.
Time recaptured per week: estimated weekly hours saved across the team based on the rebuilt workflows. Target: 5 to 15% of team capacity by week 12. Modest but real. Anything claimed above 30% is almost always fictional.
The three things that derail it
In four years of running these programs, three things consistently derail them.
First: no executive sponsor. Someone senior has to own the program publicly, attend the kickoff, and reinforce it in normal management conversations. Without that, the program becomes optional in the team's heads. Optional things get deprioritised by week 3.
Second: champions without time. Champions need four to six hours per week during weeks 5-8 to support their functional area. If those hours are not protected in their calendars by their managers, champions burn out and disengage. The program dies with them.
Third: no internal community after week 12. Programs that end cleanly at week 12 lose 60 to 80% of their measured gains within 90 days. Programs that have an ongoing prompt library, monthly internal show-and-tell, and a place to ask questions keep the gains and compound them.
- Ninety days is the shortest window to move through learn, apply, embed.
- Standardise tools in weeks 1-4 before trying to standardise workflows.
- The week 5-8 workflow rebuild is the operational spine of the program.
- Champions need four to six protected hours per week during the rebuild phase.
- Without an executive sponsor and a post-program community, gains evaporate in 90 days.
Does the team have to be 30+ people?+
No, but the economics work better above 15. Below that, a 4-week cohort or a series of workshops is usually a better fit than a full 90-day program.
What if we have a distributed team?+
The program works fully remote. The kickoff day is on Zoom, and weekly coaching and champion sessions are remote. The one element worth doing in person if possible is the champions network. Having champions meet face-to-face once during weeks 5-8 noticeably improves the embed phase.
How is the budget for this typically structured?+
Most teams treat the program as L&D budget. Some treat it as transformation budget, justified by the time-recaptured KPI. Either works. Transformation budget tends to be larger and gets approved faster against a clear ROI projection.
Can we shorten it to 60 days?+
Possible, but you cut the embed phase. The program then produces strong rebuilds but unreliable adoption. The team often drifts back to old workflows within three months. If 90 days is genuinely impossible, run a 60-day rebuild then a 30-day light-touch embed phase with a smaller weekly time commitment.
What about training the champions specifically?+
Yes, and we recommend it. A separate half-day champion training before the program starts gets them aligned on their role and gives them facilitation patterns. It is included as a configurable option in most Capability Programs.
