General Assembly stopped selling to individuals on April 28, 2026. It now trains companies only. That single fact invalidates most of what ranks for “General Assembly review,” because those ratings were left by alumni of bootcamps and short courses the company no longer runs. If you are deciding whether to put 60 employees through a General Assembly program this quarter, the 4.31-star average you found rates a different business.
Declare the interest first: SFAI Labs delivers facilitated sessions that teach teams to use AI assistants on their own documents and daily work. General Assembly is a credible alternative to us in some situations and the better choice in others, and this page says which. Everything below comes from General Assembly’s live pages, its regulatory disclosures, and independent survey data, read in October 2026.
What you are evaluating in late 2026
General Assembly was founded in 2011, grew into one of the better-known technology bootcamp brands, was acquired by the Adecco Group in 2018, and now describes itself on its employer pages as “combining strengths with LHH and as part of the Adecco Group.” Its regulatory information page carries the sentence that should govern your evaluation: “General Assembly ceased new consumer enrollments on April 28, 2026 and is concluding its consumer course operations for currently enrolled students through an orderly teach-out process.”
The company did not shut down. It changed customers. Everything it sells today is sold to an employer, which makes it a direct peer of the providers in our review of the corporate AI training market rather than a school a reader of this page would enroll in personally.
The pivot was deliberate and staged: an enterprise AI Academy launched in April 2025, four AI courses added that December, fifteen more the following January. The AI catalog is the product now, not a track inside a bootcamp catalog.
The review problem: 714 ratings of a closed business line
Course Report shows General Assembly at 4.31 out of 5 across 714 published reviews. SwitchUp shows 4.27 out of 5 across 1,005. Those are real reviews from real alumni, and they are also the most misleading input available to a team buyer. Course Report says why, in a notice at the top of its own page: “General Assembly ceased enrolling consumers on April 28, 2026. It continues to provide training for businesses. This page preserves historical consumer program information and alumni reviews.”
Read the programs those reviews attach to and the problem is plain. They rate Data Science, Full Stack Web Development, Product Management, Digital Marketing, UX Design, and Cyber Security career tracks, many of them twelve-week immersives taken by career changers paying their own tuition. Not one is a corporate AI program.
The recurring complaints follow from that context and mostly do not transfer. Career services and job placement is the dominant criticism — irrelevant when the learners already work for you. Reviewers also report instructor feedback thinning in the back half of a twelve-week course, a real risk signal for a long cohort and no signal at all for a one-day workshop.
What does transfer is the praise, because it is specific and consistent: instructors with industry experience, live delivery that holds attention, cohorts that build something. Those are properties of the delivery model, and the delivery model survived the pivot.
So the honest conclusion is narrower than either the five-star summaries or the one-star warnings. There is very little public review evidence about the business-only company, because it is roughly six months old. Treat the star ratings as evidence of teaching craft and get the rest from references, which is where our guide to verifying a training vendor’s case study earns its keep.
What a company can actually buy
Several purchases hide behind one brand name, and they fail in different ways. Separating them is most of the work.
| Product | Shape | What it is good for |
|---|---|---|
| Free one-hour classes | Generative AI Basics, Prompt Writing 101, Agentic AI Fundamentals | Testing instructor quality at no cost |
| Role workshops | One day, around seven hours — AI for Marketers, AI for Sales, AI for HR, AI for Product Managers | One job family, one sitting |
| Short workshops | Three hours — AI for Data Analysis, AI for Workplace Efficiency | A single capability, low calendar cost |
| Open cohort courses | 32 hours over 8 weeks, instructor-led, online or on campus | Deep role reskilling for a few people |
| Enterprise AI Academy | Quoted program, scoped company-wide | A staged sequence that starts with assessment |
Details read from General Assembly’s catalog and workshop pages in October 2026. Corporate training catalogs move every quarter, so confirm the list before it reaches a proposal.
The enterprise offer is a six-stage sequence whose stage names tell you what the company believes: AI Assessment (“evaluate AI readiness across teams and functions”), AI for Leaders (“equip decision-makers to guide adoption”), AI Fundamentals Training (“create shared AI fluency across the workforce”), Role-Based AI Skills Training (“move from knowledge to real-world application”), AI Champions and Agentic AI, and Scale and Sustain.
That is a coherent AI enablement program on paper, more complete than most vendors in this market publish. The add-on list beside it — offsites and hackathons, behavioral coaching, instructor augmentation, proprietary content build, content licensing — is the tell that General Assembly will also sell you material to run yourself, which matters if you are weighing building internal capability against buying it.
The cohort courses deserve a caution. For an engineer moving into AI systems work, 32 hours is the right dose. For an operations manager who needs to stop rewriting the same client email by hand, it is roughly eight times the dose required, spread across two months in which the habit either forms in week one or never forms.
What it costs, and what quote-only means
General Assembly publishes no price — not on the course catalog, not on the workshop pages, not on the enterprise pages. The $4,500 figure still circulating in search results belonged to a retired consumer product.
For reference, the market bands to expect: a private AI fluency session for a single team runs roughly $2,000 to $15,000 depending on length, group size, and how much of the material is built from your own documents. Custom automation work is a different purchase, usually $25,000 to $150,000. Those are market ranges, not General Assembly quotes, and we would not put a figure in your board paper that the vendor has not signed.
Quote-only has two consequences. The first is procurement friction — you cannot model the cost this afternoon, and whoever asked you for a number waits out a sales cycle. The second matters more: without a published rate you cannot tell whether the quote reflects a program built for you or a catalog deck relabeled with your logo. Both arrive as a PDF with your company name on the cover.
There is no free pilot tier either. The free one-hour classes are marketing sessions for individuals, so a pilot has to be negotiated into the engagement — one of several items our review of the contract terms that decide whether training lands treats as non-optional.
Who General Assembly fits
Four situations where we would tell a buyer to take the meeting:
You need role-based coverage across several job families at once. Marketing, sales, HR, product, and data each get their own material. One fluency session, ours included, does not cover five job families five different ways. If your mandate is “every department by Q2,” breadth is the requirement and General Assembly is built for it.
You are reskilling technical staff, not raising general fluency. The AI engineering and AI systems reliability tracks are for developers whose jobs are changing underneath them. That is a genuine curriculum taught over real hours, and no fluency workshop substitutes for it.
You want an assessment before you commit a budget. The AI Assessment stage exists for the buyer who cannot yet say which teams need what, and paying for that answer is defensible when the alternative is guessing across 300 people.
Your procurement process prefers a large institution. General Assembly sits inside the Adecco Group. For a CFO who has watched a training vendor fold mid-engagement, parentage is a real risk control, and the one thing a boutique provider — including us — cannot offer.
Who it does not fit
You need Monday morning to look different, not a certificate. The cohort model produces portfolio projects and verified badges — useful for the employee’s career, beside the point if your objective is that the proposal team stops spending six hours on first drafts.
Your real problem is unused licenses. The most common situation we see: the company bought AI seats, usage flattened, and someone was asked to fix it as a second job — the subject of our argument about why licenses are not capability. A general curriculum rarely moves it, because the obstacle is not knowledge of AI in the abstract. It is that nobody has applied an assistant to this company’s intake form while the people who own that form watch.
You have one team, one workflow, and no appetite for eight weeks. A 32-hour cohort is the wrong instrument for a 15-person department with one painful process. A half day on that process, with that team, beats two months of general instruction.
You need the material to be about your documents. If what you need is your month-end reconciliation, your client intake, your statement of work template, you are buying customization as an add-on to a catalog product rather than buying the thing itself.
What the cohort model gets right
Credit where it is earned, because the comparison most buyers actually face is not General Assembly against a facilitator. It is General Assembly against a self-serve video library.
Boston Consulting Group’s AI at Work research is the most useful independent evidence here. The June 2026 edition, drawn from close to 12,000 employees, managers and leaders across more than a dozen markets, found 72% say the skills expected of them have shifted while only 36% feel adequately upskilled. The 2025 edition was blunter about what closes that gap: those who receive five or more hours of training, “especially in person and with coaching,” are significantly more likely to become regular users.
A scheduled cohort delivers those hours. A library offers them and lets everyone defer. That is the strongest argument in General Assembly’s favor, and a rival provider’s own ranking of the field concedes it: a scheduled session with an instructor removes the “I’ll get to it next quarter” problem. The same arithmetic drives our comparison of a seat in a learning library against a facilitated session.
Where a session on your own documents wins
The distinction that decides most of these purchases is not live against recorded, or big vendor against small. It is generic against yours.
Mechanically: in the two weeks before a session, a facilitator doing the work asks for real artifacts — a redacted proposal, an intake form, a weekly report somebody rebuilds by hand — and builds the exercises on them. On the day the team watches an assistant handle their own material, including the places it fails, which is where the learning sits.
A catalog course cannot do that at its published shape, which is why customization is a separate line item. Ask any provider, us included, what they do in the two weeks before they arrive.
Five questions that make a quote comparable
Put these to General Assembly and every alternative, in writing.
- Who is the instructor, and what have they shipped? A name, a bio, and something they have built with AI assistants in a commercial setting. Not “a certified facilitator from our bench.”
- What percentage of the material is built from our documents? Expect a number. Zero is an honest answer and tells you what you are buying.
- Can you name a client of our size trained after April 2026? A reference from the consumer era does not describe the service you would receive.
- What happens in the thirty days after delivery? Office hours, a prompt library, a follow-up session, or nothing. Price the nothing.
- What will you show us at ninety days? Attendance and satisfaction are not outcomes.
The ninety-day test
Whichever provider you choose, decide the test before you sign and write it into the engagement. Pick three workflows someone currently does by hand and record the baseline hours. Ninety days later, check whether those workflows are being done differently, what share of trained employees used an assistant last week in the license data you already own, and whether anyone outside the trained group has copied the new method. That third signal predicts whether the capability stays once the vendor leaves.
If you want a second opinion before you commit a budget cycle, we run a free AI-readiness call — thirty minutes on where your licenses are idling and what a first session should cover. If a role-based catalog program fits you better than a facilitated session, we will say so.
Frequently asked questions
Is General Assembly still in business?
Yes, as a business-to-business training company. It ceased new consumer enrollments on April 28, 2026 and is teaching out its remaining individual students. Corporate AI training continues and is now the company’s focus.
Why are General Assembly’s reviews so mixed?
Because almost all of them rate retired consumer bootcamps rather than corporate AI programs. The dominant complaint is career services and job placement, which does not apply when the learners are your own employees. Read the ratings as evidence about teaching quality and nothing else.
How much does General Assembly AI training cost for a team?
No price is published, and every engagement starts with a quote. For context, private AI fluency sessions across the market run roughly $2,000 to $15,000 depending on length, group size, and customization. The $4,500 figure still circulating online belonged to a discontinued consumer course.
What AI courses does General Assembly offer now?
Free one-hour classes, three-hour workshops such as AI for Data Analysis and AI for Workplace Efficiency, one-day role workshops for marketing, sales, HR and product management, 32-hour instructor-led cohorts over eight weeks, and a quoted enterprise AI Academy.
Is General Assembly better than a self-paced AI course library?
For getting people to finish the training, usually yes: a scheduled cohort produces the hours survey evidence links to regular AI use, while a library depends on employees choosing to spend them. For catalog breadth at low cost per seat, the library still wins.
Does General Assembly customize training to our company?
It sells customization as add-on services — proprietary content build, content licensing, instructor augmentation and behavioral coaching — layered onto catalog programs. Ask what percentage of the material will be built from your own documents before comparing the quote against a provider that starts there.
Should we send employees to a 32-hour cohort or run a workshop?
Use a cohort when you are reskilling someone into a new role, such as a developer moving into AI systems work. Use a workshop when an existing team needs to work differently on work it already owns.
What should we ask for in a General Assembly reference call?
A client of comparable headcount and situation, trained after April 2026. Ask what percentage of the material was theirs, who taught it, and what they measured. A reference who can only describe the classroom has not answered the question.
Key takeaways
- General Assembly stopped enrolling consumers on April 28, 2026 and now trains companies only. Reviews predating that pivot rate a product you cannot buy.
- The 4.31 and 4.27 star averages come from bootcamp alumni. Course Report says so on its own page.
- Four products sit behind the brand — free classes, short and one-day workshops, 32-hour cohorts, a quoted enterprise academy — and they solve different problems.
- No price is published anywhere. Expect a sales cycle, and ask what fraction of the material is actually yours.
- It fits multi-department role coverage, technical reskilling, and buyers who need a large institutional vendor. It fits poorly when the real problem is idle licenses and one team’s workflow.
Arthur Wandzel
