Change management is a bigger determinant of program success than the technology itself. This is the machinery that makes usage stick after the consultants leave.
Enablement isn’t a phase of the program. It’s the floor it runs on.
The System
Five steps, four weeks — then wave two starts with the backlog already scored.
Office hours run weekly through weeks 5–8 — what stalled gets fixed, and wave-two ideas get scored while momentum is warm.
The Teaching Philosophy
We train the team on the buttons — when to use each skill, when not to, and what to escalate.
What most rollouts ship
An open prompt box asks everyone to become a prompt engineer. Most won’t — and quietly go back to the old way.
What we ship
Each skill has a name, an owner, a one-page doc, and a feedback channel — the workflow lives in the button.
Week One
A 60-minute kickoff, 8–12 structured 45-minute interviews, and two or three shadowing sessions — the output is a scored inventory of 30–60 processes.
Involvement is hours per person, not days — nobody gets pulled off the close to fill out a survey.
The Filter
Every inventory entry runs the same formula — the top 5–8 survive, the rest are logged as wave two.
Annualized hours × loaded rate × confidence — the dollar figure if it works.
Structured inputs, structured outputs, clear acceptance criteria — the closer, the faster it ships.
How often the pain repeats — a monthly grind beats a yearly fire drill.
Training
Each audience gets the version of the program that matches what they’re accountable for.
What to demand from the telemetry — and what to ask when it stalls.
How to review work that’s partly delegated to skills.
The buttons, hands-on with real work — max eight per session.
The Whole-Company Hour
Sixty minutes, every employee, no exceptions — built from your workflows, not stock slides.
What the models are good and bad at — and how to tell.
The data bright lines: what never gets pasted where, and why.
The skills, live — real workflows run on stage, not demos.
When not to use AI: consequential decisions, regulated topics.
How to ask for a new skill, and who your champion is.
Live, recorded, and translated where teams work in other languages — then folded into new-hire onboarding, so the program survives turnover.
The Honest Part
Every employee is quietly asking what this means for their job — so we answer early, and honestly.
Leadership states on day one what the program is for and what it isn’t. Ambiguity is what people fear.
Where AI changes a role, the description changes with it. People deserve to know what a good year looks like.
When hours come back, some fund people learning the next thing — part of the program, not an afterthought.
Teams that adopt well get recognized publicly. Adoption becomes something ambitious people want on their record.
The companies that skip this conversation get quiet resistance and empty dashboards.
The Multiplier
An existing trusted operator with protected time — peers already trust them, which is exactly why it works.
Runs the skills on real work, in front of the team.
The first stop before anything becomes a ticket.
Every skill has one. The champion reads it and acts.
Spots the next candidates and routes them to the program.
The Role, Made Concrete
Four small, repeatable habits — the difference between a program that compounds and one that coasts.
Fifteen minutes on the dashboard — drops flagged before they become trends.
Two-minute answers in the feedback channels. The fastest support desk anywhere.
Brings the top two issues, plus one wave-two candidate.
Logs what got fixed, what got asked, what’s wanted next.
A protected 10–20% of their week, formally in their goals. Never a new hire — always the operator peers already trust.
What Actually Happens
Every rollout follows the same shape: a spike, a trough, and — only if the interventions land — a habit.
Usage over time · every program, everywhere
— — dotted: the unmanaged program. It never comes back out of the trough.
Everyone tries it once. The launch numbers look wonderful and mean almost nothing.
Usage sags as novelty wears off. This is where unmanaged programs quietly die.
The skills become how the work gets done. Usage climbs past the launch spike and stays there.
Leading indicators · watched weekly
Depth, not breadth — ten people running a skill daily beats a hundred who tried it once.
A quiet channel isn’t satisfaction — it’s disengagement. Healthy skills generate requests.
The gap between first run and second — the best predictor of habitual users.
The intervention playbook
A usage drop triggers an office-hours diagnosis the same week — a fix, retraining, or retirement. Killing a dead skill is the system working, not failing.
What We Leave Running
Weekly actives per skill, invocations, hours saved against the baseline — flags wherever usage drops.
Weekly active users · by function
Est. hours saved / week · vs baseline
Top skills · this week
Flag · usage drop
Legal WAU down 9pts after a template change — routed to the champion, fixed in Thursday’s office hours.
The Honest Numbers
Weekly actives flatter everyone in week one — the honest metrics are depth per user, hours against the written baseline, and retirement rate.
weekly active at month six — after the trough, not before it
priority-team adoption where enablement ran as a system
shadow-AI usage once the sanctioned skills were the easier path
All three came from programs where enablement ran as a system, not a launch email.
After Handoff
Four roles keep it alive — one of them is the difference between a program and a memory.
Half to full-time
Often promoted from FP&A or business ops. The single point of accountability — the role most companies skip, and why rollouts stall.
10–20% each
One per function, time protected on the calendar. They keep skills honest and feedback loops short.
A few hours a month
Keeps the cadence honest and flags when model changes mean a skill should be rebuilt or retired.
From month four
Worth hiring as Company OS work starts — turns a governed skill library into infrastructure.
Post-launch support shapes
A standing weekly session — skills stay sharp and wave-two candidates keep flowing.
Lightest touch
A focused build wave each quarter. New skills ship, old ones get rebuilt as models improve.
Most common
For teams rolling straight into a Company OS — we stay until the internal team can carry it.
Company OS track
The Smallest Way In
Proof before a program: literacy, workshops, and guardrails — delivered in weeks.
Prepaid training bucket
Model tiers, data bright lines, when AI is the wrong tool.
Finance, sales & marketing, customer success, HR — each ends with five scored skill candidates.
Produces a written guardrail playbook the team owns.
flat fee · 25 hours · fully remote
Extension in five-hour blocks
Starts within a week
The cheapest way to find out what a real program would return.
Proof
Regulated industries are where enablement gets tested — the sequence holds.
Shipped across 2 functions in 60 days — where every workflow touches PHI and every skill cleared compliance first.
Underwriting cycles vs the pre-program baseline. Their governance framework became an audit asset at the next regulator review.
The plain version
The Foundation
Enablement is one of three foundation layers that run under every band.
The Rest of the Spectrum
Ready to move
One function, four weeks, and a dashboard they can argue with. If the usage isn’t there by week six, you’ll see that too — the telemetry doesn’t flatter anyone.
Talk to LightCI