claude academy: which course to take first.
anthropic put its training in one place on august 20, and it is free. the useful question is not whether to do it. it is which part, in what order, and where to stop.
claude academy is anthropic's free learning site, live since august 20, 2026. it holds two full courses, a set of short tutorials, and five product tracks. if you run a business, take ai fluency: framework and foundations first, then pick the one product track that matches the work you already do.
i am not reviewing this from the outside. i hold the completion certificates for both of the courses now sitting on that site, taken back when they lived as separate anthropic materials, plus claude 101, claude code 101, introduction to claude cowork, and introduction to agent skills. so this is less a first look and more a running order, written by someone who already did them in the wrong sequence and can tell you what that cost.
what claude academy actually is
it is a single site, academy.claude.com, that collects anthropic's courses, short tutorials, and per product learning into one place, and the content is free. per EdTech Innovation Hub (24 august 2026), the learning material does not require a sign-in at all, so you can read it before deciding anything. an account only matters if you want progress tracking and the completion badges you can put on linkedin.
reading academy.claude.com directly on 25 august 2026, the catalogue breaks into three shapes. there are two long courses. there is at least one very short tutorial, "the 4 properties of ai", listed at 7 minutes. and there are five product sections, one each for claude.ai, claude cowork, claude code, claude tag and claude platform.
that third shape is the one people get wrong. five tracks look like a menu you are supposed to finish. they are not. they are five different jobs, and you almost certainly have one.
the two courses, and which one is the real prerequisite
take ai fluency: framework and foundations first, because it is the course about deciding what to hand over, and every other course assumes you already made that decision. it runs 14 lessons and a quiz, listed at about 4 hours (academy.claude.com, read 25 august 2026).
this is the course that teaches anthropic's 4Ds: delegation, description, discernment and diligence. the framework is by rick dakan and joseph feller with anthropic, released under CC BY-NC-SA 4.0, and it is theirs, not mine. i broke the four skills down in plain business terms in the 4 skills of ai fluency. the definition the course opens with is the part worth holding onto:
working with AI in a way that is "effective, efficient, ethical and safe."
AI Fluency: Frameworks and Foundations, Dakan and Feller with Anthropic, CC BY-NC-SA 4.0
the second course, ai capabilities and limitations, runs 13 lessons and a quiz at about 3.5 hours (academy.claude.com, read 25 august 2026). it is the discernment course in practice. take it second, not because it matters less, but because it is much easier to absorb once you have a specific task in your head that you are trying to judge the output of.
roughly 7.5 hours for both. that is one working day, spread over a week if you like.
the five product tracks, and who each one is for
pick exactly one track, the one matching work you already do every week, and ignore the other four until that one has paid for itself. here is the routing, based on the five product sections listed on academy.claude.com:
| the track | who it is for | the first job it covers |
|---|---|---|
| claude.ai | anyone starting, and anyone whose whole AI use so far has been a chat box | turning a task you repeat into an ask you can describe the same way every time |
| claude cowork | non-coders who want claude working on actual files and folders, not just answering | handing over a document or spreadsheet job you currently do by hand every week |
| claude code | builders, and anyone willing to live in a terminal for an afternoon | running edits and commands across a real project instead of copying answers out of chat |
| claude tag | teams whose work already happens in slack threads | pulling claude into the thread where the decision is being made, not a separate tab |
| claude platform | developers wiring claude into a product, a client build, or an internal tool | your first api call and your first tool definition |
if two of those look right, you are probably describing two different jobs and should still start with one. which surface fits which kind of work is its own decision, and i mapped it separately in which claude surface for which task. for cowork specifically, what claude cowork actually is covers what changes when the AI can touch your files.
the order i would take them in
five steps, and the fifth is the one that decides whether the first four were worth anything.
- the 4 properties of ai, 7 minutes. cheapest possible way to find out if the site is pitched at your level.
- ai fluency: framework and foundations, about 4 hours. the 4Ds. this is the prerequisite, whatever your job is.
- ai capabilities and limitations, about 3.5 hours. discernment, applied. the skill that stops you shipping a confident wrong answer.
- one product track. from the table above. one. not five.
- one real task, run end to end, before you open a second track. your own data, your own business, start to finish.
step five is where most people quietly fail, and it does not look like failing. it looks like finishing another course. if you want a way to choose that first task well, what tasks to give ai has the test i use.
what a course cannot give you
a course teaches you the skills. it cannot know your business, and that gap is where AI adoption actually dies. about 85% of people have access to AI and about 25% use it on real work, per the IBM 2026 Global CEO Study. that 60 point spread is not a training problem in the way it looks. everyone in that gap could pass a quiz.
what the courses do not contain, because they cannot: your context. the way you already write. your offer, your client list, the eleven small rules you would tell a new hire in week one. every course output is generic until someone loads that in, and loading it in is a build, not a lesson. i wrote up why in the ai adoption gap.
so the honest read on claude academy is good, and narrower than the launch coverage suggests. it will make you fluent. it will not make you set up. those are two different days of work, and only one of them is free.
common questions
is claude academy free, and do i need an account to use it?
it is free, and per EdTech Innovation Hub (24 august 2026) the learning content does not require a sign-in, so you can read and watch before deciding anything. an account is only what lets you track progress and earn the completion badges you can add to linkedin. the site is academy.claude.com.
which claude academy course should i take first if i run a small business?
ai fluency: framework and foundations. it is 14 lessons and a quiz, about 4 hours, and it teaches the 4Ds. delegation is the part that decides whether AI works in a business, because it is about choosing which work to hand over, and that choice happens before you open the chat. do a product track second, never first.
how long does the ai fluency course actually take, and is it worth the time?
about 4 hours for framework and foundations, plus about 3.5 hours for ai capabilities and limitations, so roughly 7.5 hours for both (academy.claude.com, read 25 august 2026). it is worth it if your whole AI skill set is currently prompting, because prompting is one slice of one of the four skills. it is not worth it as a substitute for running a real task.
will a claude academy badge on my linkedin profile actually do anything for me?
a badge is a signal, not a qualification. it tells a client you have covered the vocabulary, which is worth something in a market where most people have not. it does not show you have shipped anything. what moves a conversation is a task you can describe running in your own business, so treat the badge as the receipt and the working setup as the proof.
the takeaway
do the 7 minute tutorial today. book the 4 hour framework course into one afternoon this week and the 3.5 hour capabilities course into the next. pick the single product track that matches a job you already do weekly, and then stop taking courses until that job is running with your own data in it. seven and a half hours of anthropic's training is a genuinely good deal at zero. the part that pays is the task you run afterwards.
want the shortcut, not the trial and error?
the ai builder toolkit is the set of claude skills that do the jobs in this post, ready to install and run in claude today. it is what runs after the course, when you need your own context in there. no course, no call.
see the toolkitworking through the tracks and want people to compare notes with? that is what the ai builders lounge is for.
or just follow along. new field notes most weeks on x, instagram, and tiktok.