A 30-60-90 Day AI Fluency Program Template
· 7 min read · OneRange Team
A quarter-long AI fluency program you can run without new headcount: baseline and guardrails in the first 30 days, role-specific application by 60, measured proficiency movement by 90.
Most AI fluency programs are planned as a rollout and should be planned as an experiment with a control on the calendar. The version below fits a quarter, needs no new headcount, and produces a defensible number at the end — which is what buys the budget for the next two quarters.
It assumes the definitions and capability bands in what is AI fluency, and the measurement method in how to measure AI fluency.
Before day one: choose one job family
The single most common design error is starting company-wide. Pick one job family with a real, current capability gap and a manager who wants this to work — not a volunteer pilot group of enthusiasts, whose results will not generalise to anyone.
- Between 20 and 200 people, so the result is neither anecdote nor logistics project
- A gap someone can name in a sentence, not "AI readiness" in the abstract
- A manager who will give people the hour and be visibly present in the sessions
- Work products you can actually inspect, so application evidence is available later
Days 1–30 — Ground, permit, baseline
Week 1: say why, and say what is allowed
Leadership states plainly why the organisation is investing in AI, what will and will not change about roles, and that the aim is to support people rather than replace them. Then publish the guardrails: which tools are approved, what data may never go into which tool, and who to ask when it is unclear. Uncertainty about the rules suppresses usage more than a lack of skill does, and it pushes confident people towards unapproved tools.
Weeks 2–3: baseline every person, per skill
Assess behaviour, not confidence. Each person does a small piece of their genuine work with AI assistance and is graded against a rubric, including one plausible but incorrect output planted for them to catch. Record a band per skill per person and keep the record somewhere that will still exist in ninety days.
Week 4: literacy for everyone, and read the baseline
One short universal session on capabilities, failure modes and the guardrails. Meanwhile, read the baseline properly: which skills are weakest, where the spread is widest, and which people are already Proficient — those are your champions for the next 60 days, not your training audience.
Milestone at day 30: a named owner, published guardrails, and a per-skill band for every person in the population.
Days 31–60 — Apply it to the actual job
Design against the three to five tasks that matter
For this job family, identify the handful of tasks where AI genuinely changes the work. Build training against those, using your own policies, playbooks and documentation so the guidance matches how your company works rather than a generic best practice.
Short sessions, same-day application
The rhythm that works: learn one small thing, apply it to live work the same day, return for the next. Avoid the all-day workshop — the backlog waiting afterwards makes the program itself feel like a cost. Ask people to bring a real task to every session and leave with it done.
Differentiate by band
Novice and Developing people need patterns and verification practice. Proficient people need depth — tool chaining, internal data, systematic evaluation — and should be spending part of their time coaching. Putting all four bands through identical content is the fastest way to lose the top and the bottom simultaneously.
Instrument as you go
Sample real work weekly. Note where AI-assisted output is being accepted uncritically and feed that straight back into the next session. Waiting until day 90 to look at work product wastes the only signal that tells you the program is landing.
Milestone at day 60: every person has applied AI to at least three real tasks from their own job, with feedback.
Days 61–90 — Re-measure, extend, report
Weeks 9–10: extend to a second job family
Start the same sequence with a second population, using what you learned. Running the two overlapping is what turns a pilot into a program and prevents the classic outcome where a successful pilot is celebrated and never repeated.
Weeks 11–12: re-assess and report
Re-run the original assessment — the same instrument, the same skills, the same rubric. Anything else and the delta is not interpretable. Then report three things and resist the temptation to add a fourth:
- Proficiency movement — the share of the population that moved up at least one band, by skill
- Time to competency — how long it took people to reach the target band
- Application evidence — verification catch rate and sampled quality of real AI-assisted work
Report the skills that did not move as prominently as the ones that did. A program that names its own failures gets believed on the numbers that improved.
Milestone at day 90: a before-and-after per skill for one full job family, a second family underway, and a decision about what to fund next based on evidence.
What to avoid
- Launching company-wide — you will get a completion rate and nothing else
- Skipping the baseline to save three weeks — unrecoverable; the day-90 number becomes an opinion
- One course for everyone — too basic for the people who will push the tools, too abstract for the rest
- Training away from the work — the backlog waiting afterwards turns the program into a resented cost
- Ending at day 90 — without a re-measurement cadence, a launch is all you built
How a capability platform shortens this
None of the above requires a specific vendor; it does require somewhere to hold per-skill assessments and re-run them identically. That is what OneRange Vero is for: assessments that run independently of any course against a 10,000+ skill taxonomy, delivered as quizzes, graded labs, code projects or AI role-plays, producing a proficiency level per skill per person before and after — with training generated from your own documents and assigned automatically when a score comes back low. The parts of this plan that usually consume the ninety days are the baseline and the re-assessment, and those are the parts a platform removes.
Tags: AI, Learning & Development, Playbook
FAQ
Frequently asked questions
How long does it take to build AI fluency in a team?
Moving a job family from Novice to Developing on a handful of role-specific tasks is realistic within a quarter when sessions are short and applied to real work the same day. Reaching Proficient across a whole function typically takes two to three quarters with re-measurement along the way.
What should the first 30 days of an AI fluency program cover?
Leadership context, published guardrails on approved tools and data, a behavioural baseline assessment per skill per person, and one universal literacy session. The output of the first month is a measurement, not a training milestone.
Should we pilot AI training with volunteers?
No. Volunteer groups are self-selected enthusiasts, so their results do not generalise to the population you actually need to move. Pick a job family with a real gap and a supportive manager, and include everyone in it.
How many people should be in the first cohort?
Roughly 20 to 200. Smaller and the result is anecdotal; larger and the first cohort becomes a logistics exercise before you have learned anything about what works.
What do we report at the end of 90 days?
Proficiency movement by skill, time to competency, and application evidence from sampled real work — including the skills that did not move. Completion rates belong in the operational appendix, not the headline.