Undisclosed Professional Services Firm
Professional Services / AI Adoption / Workshops & Enablement / Change Management / AI Governance
Executive brief
From 9% to 71% Weekly AI Use Across 217 People
The firm had bought AI licences for all 217 employees fourteen months earlier. Weekly active use sat at 9%, concentrated in one technology group, while unsanctioned tools spread quietly everywhere else. We designed and ran a role-based enablement program: six curricula built on real artefacts from each function, 12 hands-on cohorts, a champion network with a defined role and protected time, and a one-page acceptable-use standard with worked examples. Adoption reached 71% by week 16 and held at 68% twelve weeks after we left, because the champions — not us — had become the support model.
Weekly active AI users
9% → 71%
week 16
Employees through hands-on workshops
178 of 217
12 cohorts over 9 weeks
Paid licence utilisation
31% → 88%
week 16
Raised weekly active AI use from 9% to 71% of 217 employees over 16 weeks, still at 68% twelve weeks post workshops.
Delivered 12 hands-on workshop cohorts across six job families, built around each function’s actual work rather than generic prompt training.
Replaced an unread three-page AI policy with a one-page acceptable-use standard plus worked examples, cutting unsanctioned tool accounts from roughly 90 to 11.
Lifted paid licence utilisation from 31% to 88%, which turned an existing sunk cost into the program’s funding argument.
Engagement shape
Duration
16 wks
Phases
5
Deliverables
7
Weekly active AI users
9% → 71%
How it ran
Diagnose, rewrite the rules, build curricula, run cohorts, hand over to a champion network
A 16-week company-wide enablement program took a professional services firm from 9% to 71% weekly AI use by teaching each job family on its own work and leaving 14 trained champions behind.
Starting position
What Undisclosed Professional Services Firm was working against.
Licences had been rolled out firm-wide fourteen months earlier with an announcement, a recorded webinar, and a three-page policy. Weekly active use was 9% and almost entirely inside one technology group. Meanwhile roughly 90 unsanctioned personal AI accounts were visible in network telemetry, several handling client material. Leadership was preparing to cancel the licences on the grounds that the firm had proved it did not want them.
Constraints we worked inside
6 in playEvery one of these ruled an easier option out. They are the reason the approach looks the way it does.
Client confidentiality obligations that varied by engagement, so a single blanket rule would be either unusable or unsafe.
A billable-hours culture in which time spent learning was time not charged, making unstructured self-teaching a non-starter.
Six job families with almost nothing in common: consultants, finance, marketing, HR, legal, and technology.
Three offices across two time zones, with an earlier training program that had been delivered as a single recorded session nobody finished.
Genuine and unaddressed anxiety about whether adoption was a prelude to headcount reduction.
A prior three-page policy that partners could not summarise and therefore did not enforce.
What had to be true to finish
5 goalsAgreed up front, so the engagement could be called finished on evidence instead of opinion.
Move AI use from a technology-group habit to a firm-wide default across all six job families.
Make the acceptable-use rules short enough to remember and specific enough to apply to client work.
Build an internal support model that survives the end of the engagement.
Replace unsanctioned personal accounts with sanctioned tooling rather than by prohibition alone.
Measure adoption in a way leadership trusts, including the parts that look bad.
Industry context
Professional Services
AI Adoption
Workshops & Enablement
Change Management
AI Governance
Professional services firms sell expertise measured in hours, which makes AI adoption both unusually valuable and unusually threatening. Adoption stalls less on tooling than on three unanswered questions: whether it is permitted, whether it is safe with client material, and whether using it makes a given person look faster or look replaceable.
Approach
How the work was sequenced.
We treated low adoption as a design problem rather than a motivation problem. Every workshop used the participants’ own artefacts — real deliverables, real client contexts, redacted where required — so the first output someone produced was usable that afternoon. The champion network was built from the start as the exit strategy, not as an afterthought, with a defined role, protected hours, and an escalation path.
Delivery track
Segment width reflects the number of workstreams inside each phase — where the engagement actually spent its effort.
Phase 1 — Diagnose the stall (Weeks 1–3)
4 workstreams
Analysed 14 months of licence telemetry to find where use had taken hold, where it had been tried and abandoned, and where it had never started.
Ran 24 interviews and six focus groups, including deliberately with people who had tried AI and stopped.
Audited unsanctioned tool usage from network telemetry to understand what people were doing when the sanctioned path did not work.
Found the dominant blocker was not skill but permission ambiguity: nobody could say whether using AI on client material was allowed.
Phase 2 — Rewrite the rules (Weeks 2–5)
4 workstreams
Replaced the three-page policy with a one-page acceptable-use standard structured around data classification rather than tool names.
Wrote 18 worked examples covering the situations people actually hit, including several explicit "no" cases.
Ran the standard past risk, legal, and three engagement partners before publishing, so it could not be quietly overridden later.
Published a simple approved-tools list with a stated route for requesting additions.
Phase 3 — Build role-based curricula (Weeks 4–8)
4 workstreams
Built six curricula around each job family’s real recurring work — proposal drafting, variance analysis, campaign briefs, job descriptions, contract review, code and ticket triage.
Sourced real artefacts from each function and redacted them for workshop use, so nothing was taught on toy examples.
Designed every session as three hours, hands-on, capped at 20 people, with participants working on their own live tasks in the final hour.
Piloted with two cohorts and rewrote roughly a third of the material based on what actually failed in the room.
Phase 4 — Run the cohorts (Weeks 6–16)
4 workstreams
Delivered 12 cohorts across three offices and two time zones, blending in-person and remote delivery.
Required each participant to leave with one committed workflow change, recorded and followed up two weeks later.
Ran weekly open office hours, co-hosted by champions from week 8 and champion-led from week 12.
Grew a shared prompt and pattern library from what worked in the sessions, curated by champions rather than by us.
Phase 5 — Hand over the champion network (Weeks 10–16)
4 workstreams
Selected and trained 14 champions across all six job families and three offices, with a written role description and four protected hours a month.
Ran a train-the-trainer track so champions could deliver the core curriculum themselves.
Handed over the adoption dashboard, the curriculum, and the library to a named internal owner in the learning function.
Agreed a quarterly refresh cadence so the material does not decay as the tools change.
What the team kept
Artefacts handed over at the end of the engagement — owned and operable by Undisclosed Professional Services Firm without us.
One-page acceptable-use standard with 18 worked examples, approved by risk and legal
Six role-based curricula with facilitator guides, exercises, and redacted real-work artefacts
Train-the-trainer track and a written champion role description with protected time
Curated internal prompt and pattern library, owned by the champion network
Adoption dashboard covering weekly active use, licence utilisation, and workflow commitments
Onboarding module folded into new-joiner induction
Quarterly refresh plan with a named internal owner
Results
What actually moved.
Weekly active use reached 71% by week 16 and was still at 68% twelve weeks later, which mattered more than the peak. The clearest single lever was not the training but the rewritten policy: use rose sharply in the two weeks after the one-page standard was published, before most cohorts had run.
Outcome ledger
Every figure we measured on this engagement, including the ones that are ranges rather than headlines.
Weekly active AI users
68% at week 28, twelve weeks after the program ended.
9%
71%
Employees through hands-on workshops
Remaining staff covered by champion-led sessions and induction.
178 of 217
12 cohorts over 9 weeksPaid licence utilisation
Turned an existing sunk cost into the program’s funding argument.
31%
88%
Unsanctioned AI accounts in use
Measured from network telemetry, not self-reporting.
~90
11
Internal champions certified
Across six job families and three offices, with four protected hours a month.
14
by week 16Self-reported hours saved per user per week
Self-reported; a 40-person time study returned a lower 2.1 hours.
3.2
week 16 surveyTime from workshop to first shipped workflow change
Tracked against the commitment each participant made in-session.
11 days
medianWhat changed day to day
The part of the result that never shows up in a dashboard, but is the reason the numbers held.
Leadership reversed the decision to cancel the licences and folded the program into standard onboarding.
Client-confidentiality incidents involving AI tools went to zero over the measurement window, having previously been unmeasured because usage was invisible.
The champion network became the firm’s support model, removing the dependency on us before the engagement ended.
Adoption stopped being concentrated in the technology group — finance and marketing ended the program with the highest weekly active rates.
Partners had a clear answer for clients asking how the firm uses AI on their work, which had previously varied by whoever was asked.
Your turn
Run this play with your team.
A 16-week company-wide enablement program took a professional services firm from 9% to 71% weekly AI use by teaching each job family on its own work and leaving 14 trained champions behind. If that shape looks familiar, the first conversation is a working session, not a pitch.
AI Adoption Strategy & Program Design
Role-Based Workshop Curriculum
Hands-On Cohort Delivery
Champion Network Development
Acceptable-Use Policy & Guardrails
Internal Prompt & Pattern Library
What Undisclosed Professional Services Firm got
Weekly active AI users
9% → 71%
week 16
Employees through hands-on workshops
178 of 217
12 cohorts over 9 weeks
Paid licence utilisation
31% → 88%
week 16