How to pick an AI course as a working analyst (a 6-criteria checklist)

Three friends asked me last month which AI course they should take. All three were working analysts. All three had already started shopping. Two had already paid for courses they were quietly regretting.

I sent all three the same 6-criteria checklist. By the end of the week, two had asked for refunds from the courses they had paid for. The third was happy with their pick. None of them asked me which course I would recommend. They asked how to think about it. That is the better question.

This is the checklist. It works for any AI course aimed at working analysts, mine or anyone else's. Use it before you pay.

Criterion 1: Is the instructor a practitioner, or an instructor?

There is a real difference. A practitioner has shipped AI workflows that other humans actually use. An instructor has taught AI courses. The skill of teaching is real, but it is not the same as the skill of building things that survive contact with real stakeholders.

Check: does the instructor have a current full-time analyst or BI role? Or a recent one with shipped work you can verify? Or are they a content creator who pivoted to teaching AI in 2023?

A course built by someone whose AI work is theoretical will teach you theoretical AI. The output of theoretical AI work is unemployable.

Criterion 2: Is it live, or pre-recorded?

I will save you the rant. Pre-recorded courses have a 4% completion rate on average across the industry. Live cohort-based courses have 60-80%.

You are a working analyst. You do not have the time or willpower to push through 40 hours of video on your own. You need the calendar pressure of "the session is at 11 AM Saturday and I am paying for it whether I show up or not." That pressure is the actual product. The content is secondary.

Check: is there a fixed start date and a fixed end date? Or can you "start anytime?" Run away from the second.

Criterion 3: Do you build against a real database, or a toy one?

Most AI courses run examples against a hand-crafted CSV with 50 rows. It is clean. There are no nulls. The schema makes sense. The story is obvious.

Your actual work has nulls. The schema is half-undocumented. The story is buried under a tracking bug from Q4 that nobody remembers. If the course teaches you to write prompts against the clean toy data, you will not be able to apply them on Monday morning.

Check: what database do you build against in the course? Ideally it is at least 10,000+ rows, has nulls, has multiple tables, has the messy realities. If the answer is "we use a sample CSV we will share," it is not enough.

Criterion 4: What do you leave with: a certificate, or portable IP?

A certificate is a credential. It signals to a hiring manager that you sat through a thing. The half-life of that signal is short. Hiring managers in 2026 have seen 200 AI certificates this year.

Portable IP is the workflow you built, the prompt templates you tuned, the SQL you can paste into your own warehouse on Monday. It is yours forever. It compounds because you keep using it. It is the thing that actually changes your job.

Check: what tangible artifacts do you walk out with? n8n workflow JSON files? Prompt templates? A Slack bot pointing at your own data? If the answer is "a PDF certificate," that is a credential, not a tool.

Criterion 5: Are your cohort peers actually like you?

A course where the cohort is half college students and half senior engineers is a bad cohort for a working analyst. The pace is wrong, the questions are wrong, the breakout-room conversation is wrong.

A course where the cohort is 12 working analysts at similar career stages is the best free networking event you will attend this year. You will swap workflow JSON files. You will compare notes on which Slack bots actually got adopted at your respective companies. One of you will hire another of you in 18 months.

Check: ask the course "who is in the cohort?" If they will not tell you, or they say "varied audiences," the cohort is not curated. If they say "working BI analysts at companies with at least one data warehouse," that is curated.

Criterion 6: Does the refund + reschedule policy respect that you have a day job?

You work full time. The week before the cohort, your VP is going to drop a fire on you. You will have to skip Saturday.

A course that gives you ONE free self-serve reschedule into the next cohort respects that reality. A course that says "no refund, no reschedule, you signed up, that is your problem" is run by someone who has never had a full-time job and a learning ambition at the same time.

Check: read the refund/reschedule policy BEFORE you pay. If reschedule is not free, or not self-serve, you will end up forfeiting the money the first time something at work goes sideways.

How my two friends got refunds

Friend 1 had paid ₹35,000 for a 6-week pre-recorded course (criterion 2 fail, criterion 3 fail). She emailed the platform within their 7-day window, got the money back. Used part of it to enroll in a live weekend cohort with a practitioner instructor.

Friend 2 had paid ₹12,000 for a course where the cohort turned out to be 80% MBA students. Cohort mismatch. Got a partial refund via the platform's ombudsman. The cohort matchmaking turned out to matter more to her than she expected.

Friend 3 had picked a course that scored 6/6 on the checklist. She kept her enrollment. She was 4 weeks in when we last spoke. Loved it.

The price question

I will be direct: AI courses for working analysts in India are clustered in three price bands.

• ₹0 - ₹500: Free YouTube playlists, free MOOCs, ebooks. Good for orientation, not for portable IP. No live cohort, no instructor support, no calendar pressure.
• ₹1,500 - ₹5,000: Live weekend courses by practitioner instructors with small cohorts. The sweet spot for working analysts. Real DB, real workflows, low time commitment, ~3-12 hours total.
• ₹30,000 - ₹2 lakhs: Multi-month bootcamps. Good if you are pivoting careers full-time. Overkill if you have a job and want to add AI to your existing analyst role.

For 90% of working analysts, the ₹1,500-5,000 band is the right band. You do not need a ₹2 lakh certificate to add AI to your day job. You need 3 hours of live build time with a practitioner.

The Sprint scorecard

I will not pretend this checklist is neutral. I built a course (The 3-Hour Champion Sprint) and I think it scores well on all 6 criteria. Here is how it scores so you can verify:

• Practitioner: yes, I am Director of BI at FabHotels with shipped AI workflows that other humans actually use.
• Live, not pre-recorded: yes, every Saturday cohort is live on Zoom. There is no "watch later."
• Real database: yes, 65K rows across 4 tables (orders, users, sessions, marketing spend) in a live Postgres instance you connect to.
• Portable IP: yes, you leave with 3 n8n workflow JSON files, prompt templates, and a working Slack bot pointing at the database. All yours.
• Curated cohort: yes, the cohort is working BI/analytics professionals. I screen for this in the booking flow.
• Refund + reschedule: no cash refunds (I am direct about that), but ONE free self-serve reschedule to any future cohort, in one click, until 6 hours before your session.

Cost: ₹1,499 Early-Bird (it goes to ₹1,999 after the launch cohorts). One Saturday. Three hours. Three real workflows.

How to use this checklist on any course

Open the sales page of whatever course you are considering (including mine). Score it against the 6 criteria. Be honest. If it scores under 4/6 for your situation, pick something else.

The checklist matters more than the course. A good checklist applied to any course is a good buying decision. A bad checklist applied to the perfect course is still a bad buying decision because you cannot verify it is right.

See if the Sprint matches your 6/6

The Sprint detail page lists everything you will build, the database you will connect to, and the exact policy. Open it and check it against the 6 criteria yourself. If it scores 6/6 for your situation, the Early-Bird is ₹1,499. If it does not, pick the course that scores 6/6.