AI mentorship for professionals
AI mentorship on what you build, not what you watch.
A practitioner reviews the AI systems you build on your own work: what breaks, who gets alerted, how you roll back. Three formats, depending on where you are.
DECISION AID
Which mentorship format fits you?
Four questions, under a minute. The answer can be a format we do not sell.
Your answer appears here.
What six months cost, and who reviews your work
AI mentorship here means a practitioner reviews what you build on your own work, not a course you watch alone. If you were looking for software that mentors people with AI, that is a different product; this page is about a person reviewing your AI work.
| Format | Price | Six months | What you get |
|---|---|---|---|
| Community · our service | $49 a month or $499 a year | about $294 | Questions answered on Telegram as you go, one live session a month (or a 1-on-1, depending on member count), 250 Pitstop credits a month. |
| Expert call · our service | $499 for one hour | $499, once | You send the agenda first; the hour is prepared on your context, and you get a written report. |
| Cohort · our service | $4,999 | $4,999 | Your own workflow for six months: one mission, a one-hour group office hour and a written review each week, a build review on every system you ship, and a portfolio. |
| Course or self-study | Coursera Plus, $59 a month | about $354 | Self-paced courses. Nobody reviews what you build. |
| Bootcamp | Course Report average | $14,142 for about 14 weeks (range $3,500 to $30,000) | Full-time pace; projects are reviewed, but they are exercises rather than your own work. |
| Marketplace mentor | MentorCruise, $120 to $450 a month | $720 to $2,700 | Sessions you book with a mentor you choose; they review what you bring. |
See our three formats on the pricing page →
Who mentors you
Loïc Guyon, the operator behind LearnAIthing. Community members message him directly; cohort work is reviewed on the systems you ship.
- He builds and runs AI agents for owner-led service firms through AI Jungle, the operated-AI service of the same team.
- His site publishes four anonymised accounts of that work, each naming what he built and the decision that stayed human.
- It also logs his own errors with dates: three between July and October 2026.
SAMPLE REVIEW, WRITTEN FOR THIS PAGE TO SHOW THE FORMAT. NOT A CLIENT’S REVIEW.
“The system works on the happy path. Now show what happens when data is missing, who gets alerted, and how you roll it back.”
- System
- An agent that sorts a shared inbox and drafts replies.
- What works
- Routine requests are sorted correctly, and drafts cite the right policy.
- What breaks
- An email with two requests gets one label, and nobody is alerted when the model times out.
- Before it ships
- Split multi-request emails, alert a named person on timeout, and keep a way to turn it off.
What six months in the cohort look like
You bring one real workflow from your job or business, and the six months follow it through five steps. Each week: one mission on it, a one-hour group office hour, and a written review of what you shipped.
- 1DiagnoseMap the workflow, pick the problem, set a baseline you can measure again.
- 2BuildAssemble tools that already work; test normal, missing and odd cases.
- 3OperateRun it on real work, with a named approver and a way to stop it.
- 4EvaluateMeasure against the baseline: time, errors, cost.
- 5ImproveEnd each cycle on one decision: deploy, revise or stop.
What is proven, and what is not yet
What research says about the format, not about us: one-to-one tutoring beat classroom teaching by two standard deviations in Benjamin Bloom’s 1984 studies, and by about 0.8 of a standard deviation in Kurt VanLehn’s 2011 review of later studies. Neither measured AI mentoring, and neither measured this cohort.
- No cohort results to quote yet. The founding cohorts are what will produce them, and the evidence page says what exists today.
- One mentor, not a choice of many. If you want to pick among specialists in machine learning research or interview preparation, a marketplace fits better.
- No free trial call. The cheapest way to test the fit is one month of the Community.
- Not a bootcamp: no job placement, and the work is on business workflows rather than machine learning engineering.
Questions about AI mentorship
How is AI mentorship different from an AI course?
A course shows you tools. Mentorship here starts from a real workflow you bring: you build the system, a practitioner reviews it, and you leave with something running and a portfolio that shows it.
What do I need to bring?
A real workflow you understand, from your job or your business. The cohort is built on it: you build, test, document and defend your own system.
Can I try it before the cohort?
There is no free trial call. One month of the Community or a single prepared expert call is the low-cost way to see how the reviews work before you apply.
What is a Deployed Business Operator?
Someone who diagnoses, builds, operates, evaluates and improves AI systems on real business work, and can defend each choice. The cohort is built to make you one.
What are Pitstop credits?
Credits to run agent work on Pitstop, the workspace where agents ask before anything leaves. Community members get 250 a month and use them without building that rail themselves.
Apply for an individual cohort
The cohort runs for six months; dates for the next cohort are on request. Tell us which workflow you would bring and we reply within two business days.