Type “AI courses” into Google and you get a wall of near-identical landing pages. Hands-on projects. Industry experts. Certificate on completion. The details that actually decide whether the thing is worth your money sit three clicks deeper: who teaches it, what you build, and whether anyone doing the hiring has heard of it.
So this isn’t a round-up of every option out there. It’s a way to sort them, plus honest notes on what the bigger platforms actually deliver.
What “AI course” means now
The term has stretched to cover at least four different products. A weekend workshop on writing better prompts for a marketing team. A twelve-week Python and scikit-learn foundation. A specialisation on transformer architecture that assumes you already know calculus. And applied courses aimed at designers, analysts or product managers who want to use AI tools well rather than build models from scratch.
All four get filed under the same keyword. Before you compare prices, work out which one you need. Someone trying to automate reporting at work has almost nothing in common with someone preparing for a machine learning engineer interview.
The four formats, and who each one suits
- Self-paced video libraries. Cheap, flexible, easy to abandon halfway through. Best if you already know how you learn and can hold yourself to a schedule.
- University-backed specialisations. Solid theory, slower pace, strong on fundamentals. Good for filling genuine gaps in maths or statistics.
- Cohort-based bootcamps. Expensive, intense, deadline-driven. Worth it when you need accountability and a finished portfolio at the end.
- Live practitioner workshops. Narrow and current. Useful for people already working in a field who need to catch up on specific tooling fast.
Format matters more than brand. A £50 self-paced course you finish beats a £2,000 bootcamp you drop in week three.
What to check before you hand over a card
Does the instructor still build things?
Look them up. A course taught by someone shipping production systems last year reads differently from one assembled out of textbook chapters. Check the dates on the syllabus too. Anything citing GPT-3 as the state of the art was written a long time ago in AI years.
What do you actually walk away with?
“Build a chatbot” is not a portfolio piece. “Fine-tune a small model on 4,000 customer support tickets and measure the accuracy gain” is. Ask to see the real project briefs before enrolling. If the sales page won’t show them, that tells you something.
Are you paying for teaching or for credentials?
Part of the fee buys mentorship, feedback and career help. Part buys a logo for your LinkedIn profile. Both can be legitimate reasons to enrol, but you should know which one is doing the heavy lifting for you.
Free routes that are genuinely worth the time
You can cover a surprising amount of ground without spending anything. Alison’s free AI courses are short, certificate-backed and good for building vocabulary before you commit to anything paid. They won’t turn you into an engineer, but they’ll stop you wasting money on an advanced course you can’t follow. If you want something with a bit more academic weight, FutureLearn’s AI catalogue mixes university-produced material with shorter industry courses, and the quality varies enough that it’s worth reading reviews before you press start.
The catch with free is completion. Most people who sign up never finish. Paid courses only do better because you can feel the money leaving your account.
Paid programmes: what you’re really buying
At the bootcamp end, the differences are mostly structure and support rather than content. BrainStation’s AI courses suit people who want a sprint-based curriculum with a cohort alongside them. The trade-off is the price tag and the fact that it assumes a serious weekly time commitment.
If you’re further along and want to learn from people running AI products right now, Maven’s live courses are built around short, intense sprints with practitioners rather than a semester-long arc. For designers specifically, Designlab’s mentor-led AI track is one of the few that treats AI as a design material instead of a coding exercise bolted onto a UX syllabus.
Should you specialise early?
General “learn AI” courses give you a map. Specialising gets you hired. Computer vision is a good example: image pipelines, model training, deployment on edge devices. It rewards focused study rather than broad sampling. OpenCV-based AI courses are the usual entry point, and they’re a fair test of whether you enjoy the work before you commit to a longer path.
Pick one area by week six. Not permanently, just enough to go deep somewhere rather than shallow everywhere.
A ninety-day plan that survives contact with real life
Most people over-plan and under-execute. A simpler version:
- Weeks 1-3: Python basics plus one free intro course. One hour a day, protected in your calendar.
- Weeks 4-7: Pick a specialism and work through a structured course. Start the GitHub repo in the first week.
- Weeks 8-10: Build one project end to end, including the parts nobody films for YouTube: cleaning data, deploying it, evaluating whether it works.
- Weeks 11-13: Write the project up, apply to five roles, and ask someone working in the field to give you honest feedback on the write-up.
Two projects you understand deeply beat six you copied from a tutorial and couldn’t explain under pressure.
How to tell by week two whether it’s working
You should be able to explain, out loud and without notes, what you built last week and why it worked. If you can’t, either the course is moving too fast or you’re watching rather than doing. That one test catches more bad enrolments than any review section will.
The other signal is the difference between boredom and confusion. Confusion is fine, because it means you’re at the edge of what you know, which is where learning happens. Boredom means the course is pitched below you, and no certificate is worth eight weeks of coasting.
One last piece of practical advice. Before you buy anything, message a current student or a recent graduate on LinkedIn and ask two questions: what did you build, and would you pay for it again? The answers arrive within a day and they’re worth more than any course landing page.