Split 90 days into three blocks of about four weeks, each ending in a checkpoint that produces evidence: first the basics and a badge, then hands-on labs and review practice, then an assessed certificate or a portfolio of decisions. Plan three to four hours a week. The template sets out what to study, do and show each week.
Why plan by evidence instead of by course?
Because employers will not ask which courses you watched. They will ask what you can do and how they can check it. A plan that ends each block with something checkable, such as a badge, a decision log or a certificate, gives you material for your CV and interviews at day 30, 60 and 90, not just at the end.
It also keeps you honest about what "AI skills" means for most jobs. You do not need to learn to build models. Most roles opening up around agents need people who can direct an agent, review what it proposes and spot when it is wrong. The guide on whether prompt engineering is still worth learning explains why review and oversight skills last longer than prompting tricks.
How do you fill it in?
Start with your end goal in one line: "Review agent decisions in a claims team" or "Move into an AI operations analyst role". Then fill each block with modules that serve that goal. Book fixed study times; a learning streak of short, regular sessions holds up better than weekend marathons.
Put hands-on work in every block. Reading about held decisions is not the same as clearing a queue of them. If your employer offers a learning credit or study time, write it into the plan and show your manager.
The 90-day plan template
| Weeks | Focus | Study | Hands-on | Checkpoint evidence |
|---|---|---|---|---|
| 1 to 2 | Why agents need governing | [introductory modules] | [read three real decision records, or samples] | Notes in your own words |
| 3 to 4 | How agents decide and are limited | [modules on depth, verdicts, sources] | [first lab] | Day 30: [introductory badge or equivalent] |
| 5 to 6 | Human review | [review modules] | [review labs: approve, amend, reject, escalate] | Decision log started |
| 7 to 8 | When things go wrong; keeping knowledge current | [modules] | [labs on errors and out-of-date sources] | Day 60: decision log of [number] cases with reasons; module tests passed |
| 9 to 10 | Exam preparation | [revision of weakest topics] | [timed practice] | Written exam taken |
| 11 to 13 | Assessed practice | [—] | [practicum or portfolio project] | Day 90: [certificate or portfolio], CV and LinkedIn updated |
How should graduates and freelancers adapt it?
Graduates usually have more hours but less work context. Spend the extra time on labs and on learning one industry's rules properly, because you cannot draw on a job's experience in interviews. The 90-day plan for graduates goes further.
Freelancers often need evidence clients can see before a full certificate. Bring the decision log forward and keep it presentable from week five. The 90-day plan for freelancers shows how to turn it into a client-facing summary.
A filled sample
A data entry clerk we will call Ben, at an invented logistics firm, wants to move into review work. He has about four hours a week.
| Weeks | Focus | What Ben does | Checkpoint |
|---|---|---|---|
| 1 to 4 | Foundations | Six introductory modules on his commute, as audio; one lab on reading a decision record | Day 30: Foundations badge |
| 5 to 8 | Operator modules one to six | Two evenings a week; review labs on a free workspace | Day 60: log of 60 practice decisions, each with a reason; module tests passed |
| 9 to 10 | Exam preparation | Revises usage and billing, his weakest area; timed run of 40 questions | Exam passed |
| 11 to 13 | Practicum | Two weeks on a simulated queue, every morning before work | Day 90: operator certificate; CV and LinkedIn updated |
Ben's plan is tight, and he knows it. If the exam slips, he moves the practicum a fortnight later rather than squeezing it, because the practicum allows no missed red line.
Where the Academy fits
The Pryme Intelligence Academy tracks map onto this plan cleanly: Foundations, about four hours and free, fills the first block; the Operator track, about 14 hours plus a two-week practicum, fills the second and third. Lessons, labs and module tests are free, and labs run in a free Build workspace.
Questions people ask
Can I do this in less than 90 days?
Yes, with more hours a week, but keep the checkpoints. The point is evidence at each stage, not speed.
What if I only want the basics?
Stop after day 30. A free introductory badge and a clear understanding of how agents are governed is a sensible place to pause, especially if your employer has not started using agents yet.
How do I show this to employers?
Put the checkpoint evidence on your CV and link anything that can be checked. The AI skills for job seekers guide explains what employers ask you to prove.