Most second years are the first year again at a slightly better rate. That is where an AI enablement renewal drifts when nobody redesigns the scope, and it is the most expensive way to spend a training budget twice.
A second year is a different purchase. Year one teaches staff to work with ChatGPT, Claude, Gemini, or Microsoft Copilot on their own documents. Year two buys the conditions under which that holds up once the sessions stop and the quarter gets busy.
Two different products, so not one scope, price shape, or owner.
What follows: what belongs in year two, what to cut, what it costs, and when to stop renewing.
What a second year is for
Year one has one job: get a named set of people fluent enough with a model to do real work in front of a facilitator. Year two has to make that fluency survive turnover, a reorganization, and the two weeks in March when nobody has time for anything.
Three things are true at month twelve of almost every corporate AI enablement engagement we see, and they set the agenda for year two.
- Some people got good and most got exposed. The group that supplied real documents now works differently. The rest sat through a session. Treating both the same wastes money on one and fails the other.
- New hires arrived with no coverage. Anyone onboarded during year one missed the program, and by month twenty-four that group is large enough to dilute what you built.
- The hard workflows are untouched. Early sessions take the easy wins: drafts, summaries, first-pass research. The work that holds real hours is messier, more regulated, or owned by someone who was not in the room.
The enablement playbook that runs from license rollout to measurable capability lays out the first-year arc. Year two is the part most buyers never scope, because the proposal in front of them copies last year’s.
What the evidence says year two should buy
The training-transfer research answers this more clearly than any vendor deck. A 2020 meta-analysis in Human Factors by Hughes, Zajac, Woods, and Salas tested how the work environment affects whether training transfers and whether it is still in use later.
Peer, supervisor, and organizational support together explained about 32% of the variance in transfer, and peer support accounted for most of that (Hughes et al., 2020). All three predicted sustainment, with peer and supervisor support strongest.
Read that as a shopping list. What keeps a capability alive is not instruction. It is colleagues using the tool in the open and managers who expect the new method in the work they review.
So year two buys peer and manager support first, teaching second: funded time for internal practitioners, short clinics inside real work instead of standalone sessions, and a half-day with line managers on what to ask for when they review output. Most renewal proposals invert this and sell another round of cohorts.
Five line items that belong in a second year
| Line item | What it buys | Why year two and not year one |
|---|---|---|
| Practitioner development | Two to five internal people trained to run clinics and review work | Needs people who are already fluent, which only exists after a first year |
| Manager expectations session | Half a day with the people who approve the work | Managers set the standard the method has to meet; sessions alone cannot |
| Workflow deep dives | Three to five hard, named workflows rebuilt end to end | The messy ones get deferred in year one and hold the real hours |
| Onboarding coverage | A repeatable session for every new hire, run by you | Dilution is the quiet way a program dies in its second year |
| A recheck with teeth | A measured before-and-after on the workflows you claimed | A second year with no measure is a subscription, not a program |
The first two are the ones buyers skip and the research points at. The fourth is the one a CFO grasps at once: you are paying to stop the asset decaying. On the fifth, the measurement guide on proving capability moved sets out which baselines hold up under a finance review and which collapse.
What to cut from an AI enablement renewal
Most renewal proposals we have read add scope. A good second year is smaller than the first in at least three places.
The all-hands session. Right once, to set a floor and a policy. Running it again teaches the people who already know and the ones who tuned out.
Catalog seats nobody opened. Pull the usage export. If a library sat under 15% active use, that line is a donation unless a named role needs a specific track.
Facilitator hours that duplicate your own people. With two internal practitioners in place, half the clinic hours should move to them and the outside facilitator should review rather than deliver. That is a cost cut and a capability gain in one move.
Advisory hours with no deliverable. Attach one, or drop it.
Spend the savings on depth: fewer rooms, longer sessions, harder workflows, and the practitioner time that makes year three cheap.
What a second year of AI enablement should cost
A well-designed second year should cost less in total and more per head than the first: fewer people, more hours each, plus a block of internal practitioner time.
AI enablement services are priced in a wide band. Workshop-based work runs roughly $2,000 to $15,000 per engagement, by cohort size and session count. Advisory retainers are posted publicly too: Enablify’s tiers run $4,000, $10,000, and $25,000 a year, where the middle one buys eight advisory calls, a readiness check, a governance review, and a single half-day department workshop (published tiers, read October 2026).
Read that shape closely. Eight calls and one session buys access to opinions, not supervised practice, and it will not move a workflow.
Budget context matters when you take this to finance. ATD’s 2026 State of the Industry puts direct learning spend at $846 per employee for 2025, down from $1,254, while formal learning hours rose to 16.7 from 13.7, the first increase in five years (ATD).
More hours, less money per head. A second-year ask lands in that climate, which is why it has to be narrower and better evidenced than the first.
Structure decides more than the number. The comparison of per-seat, per-cohort, and retainer pricing shows which shapes fund scheduled practice and which only fund entry. A year-two retainer earns its keep when it covers practitioner supervision and onboarding on a standing basis, and stops earning when it becomes a monthly call.
Who owns the program by month 13
By the start of year two the program has a named internal owner with real hours, or a well-liked volunteer. That one fact predicts whether year three is a program or a line item nobody defends. The owner needs four things, and a renewal is the moment to get them in writing:
- Hours, in writing. A number per week, taken off something else. These jobs usually land on a capable person as a second job, which is the most common reason a second year underdelivers.
- Standing to change a workflow. Someone who can tell a function to stop doing a task the old way, or borrow that authority from a sponsor.
- The artifacts. Editable exercise sets, prompt libraries, job aids, recordings, baseline data. The guide on what you should receive at the end of an engagement lists what to demand and when to ask.
- A line to the facilitator. Review time, not delivery time. An hour a month reading what internal practitioners produced beats another cohort.
If the year-two proposal does not name your owner and define what moves to them, say so in the meeting and watch which way the scope goes.
The day a week nobody reinvested
BCG’s June 2026 workforce study found that 42% of regular AI users among frontline employees save about eight hours a week. Of those, 66% report limited or no guidance on what to do with that time, and more than half do not redirect it into higher-value work (BCG).
That is the central finding for anyone pricing a second year. Year one, done well, produced time. Nobody said where to put it, so it went into inbox, meetings, and longer versions of the same tasks.
The gap between hours saved and hours redeployed is where the return on year one sits, unclaimed.
Fixing it is management work with a training component. Pick two roles, then name the task each should stop and the task each should start.
Put it in the manager’s review standard and use the retainer to supervise the switch for a quarter. No vendor catalog lists this, because it is not a course.
The wider argument runs through the case against mistaking licenses for capability.
A twelve-month shape for year two
A second year that works tends to look like this, and dates matter more than the sequence.
Q1: recheck and narrow. Re-measure the workflows year one claimed. Pick this year’s three to five hard ones. Train two to five internal practitioners. Run the manager session.
Q2: deep work. Clinics inside real workflows, led by internal practitioners with the facilitator reviewing. Your onboarding session runs for the first time.
Q3: the reinvestment quarter. Two roles, one task stopped and one started each, held in the manager’s review standard. Second onboarding cohort. Mid-year measure, written down whether or not it flatters anyone.
Q4: handover and the year-three call. Artifacts audited and in your hands, owner’s hours confirmed, then the decision: a third year, a smaller maintenance deal, or nothing.
The program timeline guide covers how long each stage takes before it shows up in work product.
When to let the retainer end
Three signals say the renewal should not happen. None is about price.
The proposal is last year’s with a new date. Same cohorts, same curriculum, same deliverables. A partner who sat in your rooms for twelve months and has no sharper plan than the one they sold cold has told you what they know.
Nobody can name what year two will change. Not a theme: a workflow and a measure. If the answer is “continued support”, the engagement has become a subscription, and the signals in the guide to a contract that has stopped working are worth reading before you sign.
Your side still has no owner with hours. Then a second year buys the same outcome as the first. Fix the ownership and come back in a quarter. Renewing without it is the most reliable way to pay twice for one capability.
Letting the retainer end is not the same as ending the program. Keep the artifacts and the practitioners, run a quarter yourself, and shop again knowing what you need.
If you have EU operations, Article 4 of the EU AI Act, applicable since 2 February 2025, requires deployers to ensure a sufficient level of AI literacy among staff who work with AI systems (Article 4 text). That duty sits with you and does not expire when a supplier does, so onboarding coverage has to continue either way.
Frequently asked questions
What should a second year of AI enablement include?
Internal practitioner development, a session with line managers, three to five hard workflows rebuilt end to end, onboarding coverage you run, and a measured recheck. Teaching stays, aimed at fewer people for longer.
Should year two cost more or less than year one?
Less in total, more per participant. You are paying for depth with a narrower group plus internal practitioner time, and dropping the all-hands session and unused catalog seats that carried the first-year number.
We trained everyone already. What is left to buy?
Sustainment and the hard workflows. The people who supplied real documents work differently; the rest sat through a session, and the workflows holding your real hours were deferred for being messy.
How do I justify an AI enablement renewal to a CFO?
Bring a before-and-after on named workflows, the hours saved and where they went, the onboarding gap, and a scope smaller than year one. A renewal argued on tool access loses, because access is already solved.
Who should own the program in year two?
A named person with written hours taken off something else, plus standing to change how a function works. A proposal that does not name that person is built to keep you dependent.
What should we cut from the scope at renewal?
The all-hands session, any catalog seats under about 15% active use, facilitator hours that duplicate internal practitioners, and advisory time with no deliverable attached. Spend the savings on longer sessions with fewer people.
Is a retainer better than a fixed second-year project?
A retainer suits year two when it funds standing practitioner supervision and onboarding coverage. It is the wrong shape when it collapses into a monthly call, which is what most posted advisory tiers amount to.
How often should new hires get AI enablement?
Build one repeatable session into onboarding and run it monthly or quarterly by hiring pace. Uncovered new hires are how a second year quietly undoes the first.
When should we not renew an AI enablement retainer?
When the proposal repeats last year, when nobody can name a workflow and a measure the second year will change, or when your side still has no owner with hours. The first two are the vendor’s problem and the third is yours.
What happens in year three?
Either a smaller arrangement covering onboarding and a periodic review, or nothing, because the capability now runs internally. A third year that looks like the second means the handover never happened.
Key takeaways
- Year two is a different product. Teaching buys fluency; the second year buys the conditions that keep it.
- The transfer research points at peers and managers, not instruction: peer support explained most of the variance in whether training was still in use later.
- Five line items earn a second year: internal practitioners, a manager session, hard workflows, onboarding coverage, and a real recheck.
- Cut the all-hands session, dormant seats, duplicated facilitator hours, and advisory time with no deliverable.
- Expect to pay less in total and more per head. Market workshop work runs about $2,000 to $15,000, and posted advisory retainers $4,000 to $25,000 a year.
- Hours saved in year one were mostly never redeployed. Naming what two roles stop and start is the highest-value thing year two can hold.
- No internal owner with written hours means no renewal yet. Fix that first, then buy.
If a second-year proposal is on your desk and it looks like the first one, bring it to a free AI-readiness call. We will read the scope line by line, say which items the evidence supports, and tell you what should come out before you sign.
Arthur Wandzel
