Sheet 08
Enablement
Access is not adoption. Handing 1,240 people a license and a policy produces a small group who were already going to use these tools and a large group who conclude the whole thing is not for them. Enablement is the work of changing that second number, and it is mostly not a training problem.
Three pathways, by what the role requires
A manager who has to review someone else’s AI-assisted work needs something an individual contributor does not. A person building a prompt twenty colleagues will rely on needs more again. One curriculum for everyone serves none of them well.
General employee
Anyone with an approved tool license who is not managing people or building workflows
Most people do not want to learn about AI. They want to finish a task without getting themselves or the firm in trouble. The pathway is built around that.
Modules
| Module | Length | Format | What someone can do afterwards |
|---|---|---|---|
| M-01What these tools are good and bad at | 20 min | Recorded, watch anytime | Can name three tasks worth trying and three where the tool will confidently produce something wrong. |
| M-02What may go in, and what may not | 20 min | Recorded, with a one-page reference | Can classify the material in front of them and knows the specific answer for client information. |
| M-03Getting a useful answer | 30 min | Live, hands-on with the participant's own work | Can structure a request with context, constraints, and an example, and knows how to iterate rather than starting over. |
| M-04You own the output | 20 min | Recorded | Understands that review is not optional, knows what to check for, and knows where to report a problem. |
Reinforcement
A session nobody returns to is entertainment, not training.
- Weekly office hours, open to anyone, no agenda required
- One-page quick references linked from the tool itself
- Monthly two-minute email with one practical example from a colleague
How completion is judged
Uses an approved tool at least weekly and can state what they may not put into it.
Office hours
Every Wednesday, 45 minutes, open link, no agenda required
The most useful thing the program does most weeks. People bring a real task, get unstuck in ten minutes, and the program learns what the guidance failed to explain.
How the 45 minutes runs
- First 10 minutes: anything anyone brings
- Middle: one worked example, usually from a question asked twice in the past month
- Last 5 minutes: what changed in approved tools or guidance
What the program takes away from it
- Questions asked more than twice become a quick reference
- Confusion that traces to policy wording goes back to the policy owner
- Requests that belong in intake get walked to the form during the session
The champion network
One person per office or practice group who is visibly using the tools and is willing to be asked. Not an expert — a first stop who knows when to escalate.
Commitment: Roughly two hours a month
What a champion does
- Answer or route colleagues' questions locally
- Bring recurring local confusion to the monthly roundtable
- Pilot new guidance before it is published firmwide
- Flag unapproved tool use as a conversation, not a report
What a champion gets
- Early access to evaluate new capability
- Direct line to the program and to functional reviewers
- Named credit in program reporting to leadership
What a champion is explicitly not
Written down and said out loud when someone joins. Volunteer roles collapse when the scope quietly expands.
- Approving anything
- Security or privacy determinations
- Being available on demand
Training calendar
6 sessions delivered, 2 scheduled. 80% of registrations turned into attendance, averaging 4.4 out of 5. Next up is Recruiting pilot: tool scope and oversight on 10 Sep 2026. Show rate is tracked because a session people register for and skip is telling you something about their week, not their interest.
| ID | Date | Session | Audience | Format | Facilitator | Registered | Attended | Feedback |
|---|---|---|---|---|---|---|---|---|
| TR-01 | 04 Jun 2026 | AI fundamentals for everyone | All employees | Live virtual, recorded | AI Program Manager | 214 | 168 | 4.3 / 5 |
| TR-02 | 18 Jun 2026 | Data handling and classification in practice | All employees | Live virtual with IT & Security | AI Program Manager with Marcus Dell | 186 | 141 | 4.5 / 5 |
| TR-03 | 09 Jul 2026 | Prompting fundamentals, bring your own work | All employees | Live hands-on, capped at 30 | AI champions | 30 | 27 | 4.7 / 5 |
| TR-04 | 23 Jul 2026 | Reviewing AI-assisted work | Managers and project managers | Live workshop | AI Program Manager with Alan Petrov | 62 | 51 | 4.1 / 5 |
| TR-05 | 06 Aug 2026 | Approved-tool onboarding | New license recipients | Live virtual, monthly | IT & Security | 48 | 44 | 4.2 / 5 |
| TR-06 | 20 Aug 2026 | Champion roundtable: what is not working | AI champion network | Live discussion | AI Program Manager | 18 | 16 | 4.6 / 5 |
| TR-07 | 10 Sep 2026 | Recruiting pilot: tool scope and oversight | Recruiters and participating hiring managers | Live workshop, required before pilot use | People & Culture with AI Program Manager | 26 | Scheduled | — |
| TR-08 | 24 Sep 2026 | Prompting fundamentals, session 4 | All employees | Live hands-on, capped at 30 | AI champions | 24 | Scheduled | — |
The adoption journey
Five stages, each needing something different from the program. Most enablement effort goes into the first two stages and most of the value appears in the last two, which is worth being deliberate about.
| Stage | Where the person actually is | What the program does | Signal they have moved on |
|---|---|---|---|
| 01Aware | Has heard the firm has approved tools and a program, but has not used anything. | Monthly two-minute email with one concrete example; manager briefings. | Can name where to go with a question. |
| 02Curious | Wants to try something but is unsure what is allowed. | AI fundamentals module and the one-page data reference; office hours. | Attends a session or shows up at office hours. |
| 03First use | Tries it on one real task and gets a mediocre result. | Hands-on prompting session using their own work, not a generic exercise. | First successful task they would repeat. |
| 04Habitual | Uses it weekly for a few known tasks and has developed personal habits. | Quick references, colleague examples, champion nearby. | Weekly active use sustained over a month. |
| 05Contributing | Shares prompts, answers colleagues' questions, spots gaps. | Champion network, prompt register, early access to new capability. | Contributes a documented prompt or brings a use case to intake. |
Questions people actually ask
Each one carries where it came from. A FAQ written by guessing at questions is a marketing document; a FAQ written from what people asked at office hours is a record of where the guidance failed.
Can I put client information into an approved tool?
Not by default. Client agreements differ, and some restrict processing project material in third-party services. Bring the specific client and material to intake and Risk will answer for that agreement. The answer often is yes, with conditions.
Source: Asked at office hours 11 times since June — the most common question in the program.
Do I have to say when I used AI to help write something?
For internal work, no. For anything published externally or submitted to a client, follow the disclosure practice — which is being finalized. Until it is adopted, public content goes through Marketing with a named author who verified the content.
Source: Open action ACT-042; exception EXC-01 covers the interim.
The tool gave me something wrong. Is that a problem I should report?
Report it if it nearly reached a client, a deliverable, or a decision. That is how the program learns where review is doing real work. Wrong output caught in review is the system functioning, not a failure.
Source: Standing guidance, reinforced in the review workshop.
Why does my request need a review when my colleague's did not?
Three things usually explain it: whether the output leaves the firm, what data is involved, and whether the tool takes action rather than suggesting. The triage page shows exactly which answers moved a request into a longer path.
Source: Recurring question; answered by publishing the triage model rather than case by case.
How long will my request take?
Standard requests are answered in the weekly triage, usually within five business days. Structured review takes two to three weeks. Elevated review takes four to six, longer when evidence a reviewer needs does not exist yet.
Source: Published pathway timelines.
I found a tool that would help my team. Can I just expense it?
Bring it to intake first. It is a short form, and it takes about ten minutes. A tool that is already in use is harder to review honestly, and unwinding it later costs your team more than the wait.
Source: Prompted by three shadow subscriptions found during discovery.
Is the firm going to use this to reduce headcount?
That is a leadership question, not a program question, and it deserves a direct answer from leadership rather than a reassuring one from a program manager. What the program can say: no pilot in the portfolio has a headcount reduction as its stated objective, and every pilot's measures are published.
Source: Asked in every listening session. Routed to leadership rather than answered around.
Can I build something with a tool that is not on the approved list?
Not for firm work. If the approved tools cannot do it, that is useful information — bring the problem to intake and the gap gets recorded. Several approved capabilities exist because someone reported a gap instead of working around it.
Source: Standing guidance.
How a question becomes guidance
Collecting feedback is easy and mostly worthless on its own. The steps that matter are diagnosing what kind of gap it is, and telling the people who asked what changed.
- 01
Capture
Office hours, training feedback forms, champion roundtable, intake conversations, and pilot check-ins. Every channel logs the question, not just the answer.
- 02
Cluster
Monthly, questions are grouped by theme. Anything asked more than twice is a candidate for published guidance.
- 03
Diagnose
Is this a knowledge gap, a guidance gap, a tooling gap, or a policy problem? Each has a different owner, and mislabeling wastes everyone's time.
- 04
Publish or escalate
Knowledge and guidance gaps become a one-page reference reviewed by the functional owner. Tooling and policy problems go to the governance committee with the evidence attached.
- 05
Close the loop
The people who asked get told what changed. This is the step programs skip, and it is why people stop reporting things.
The barriers this loop has surfaced, and what was done about each