The quote you approved is the smallest number in the transaction. A half-day session for 25 managers is 100 person-hours, and at the Bureau of Labor Statistics figure for management and professional compensation — $78.85 per hour worked in June 2026 — that room costs about $7,900 in payroll before anyone sends an invoice. A vendor fee in the usual market range for hands-on AI training, roughly $2,000 to $15,000, sits beside that number rather than above it.
So the question to ask about a generic slide deck is not whether it was cheap. It is what else that $7,900 of attention was going to buy, and whether the deck gave it back.
What the room costs while the deck is on screen
Run the arithmetic on any corporate AI workshop you are considering: headcount, times contact hours, times a loaded hourly rate. Use the BLS Employer Costs for Employee Compensation series rather than salary divided by 2,080, because the employer figure includes benefits and is what your company actually spends. For June 2026, management, professional and related occupations came to $78.85 per hour worked in total compensation, of which $53.15 was wages and salaries.
| Format | Headcount | Person-hours | Compensation in the room |
|---|---|---|---|
| 90 minutes | 12 | 18 | ~$1,420 |
| Half day | 25 | 100 | ~$7,900 |
| Full day | 25 | 175 | ~$13,800 |
| Full day | 60 | 420 | ~$33,100 |
The last row gets approved most easily, because a single large session looks efficient on a calendar. It is the most expensive line here, and the format in which a generic deck does the least good, since nobody practices on their own work in a room of sixty.
There is a second budget line worth comparing against. The Association for Talent Development’s 2025 State of the Industry report put average direct learning expenditure at $1,254 per employee across 13.7 formal learning hours used per employee in 2024, so a four-hour session consumes close to 30% of a typical employee’s formal learning hours for the year. Divide those two figures and direct spend runs about $92 per learning hour — our arithmetic on their numbers, not one they publish — against $78.85 of compensation in the chair for the same hour. Most budgets track only one of the two.
The pre-work request is the tell
Every critique of generic AI sessions uses the word “generic” as an accusation and then declines to define it. There is a definition, and it is checkable before you sign: a session built on your team’s work has to ask for your team’s work.
So look at the pre-work email — not the agenda or the learning objectives, but the logistics note telling participants what to bring.
BrainStation publishes a complete AI Fundamentals curriculum: a full-day in-person format running 9:00 to 4:30, in three parts. Part one is AI in Plain Language, Meet the Tools, and Ethics and Responsible AI. Part two is prompting, part three agents and workflows. The page says exercises are grounded in real work and that each attendee leaves with a personalized action plan. The “what to bring” list is a laptop, a charger and an optional notebook.
That is not a criticism of BrainStation, which runs a reputable program and is more transparent about its curriculum than most. It is the clearest published example of the structural point: a session meant to work on your documents has to collect your documents, and the request has to go out before the day. When nothing is collected, “real work” means real-looking work, and the material is interchangeable across every company booking the same date. Our walk-through of what a team must do before the trainer arrives covers what that request should contain.
The two-line test to run on any proposal: what are you asking each participant to bring, and by when; and what will you do differently in the room if they bring it. A vendor with an answer to both has built something for you. A vendor answering only the first has a form. A vendor with neither has a deck, and the deck costs the same $7,900 of attention as anything else.
Where the spend leaks
A generic session is rarely bad in the room. People enjoy it and satisfaction scores come back fine. The loss happens afterward, and its size is one of the better-measured findings in the literature.
Arthur, Bennett, Edens and Bell pooled 1,152 effect sizes from 165 sources on organizational training. Mean effects were solidly positive: d = 0.63 against learning criteria across 234 data points, d = 0.62 against behavioral criteria. Training works. The finding that matters to a budget is their within-study comparison, among studies measuring both learning and behavior on the same people: the average decrease from the learning measure to the behavioral measure was 0.77, which the authors call a fairly large decrease and attribute to transfer conditions and skill loss.
Read that as an accounting statement. You can buy the learning reliably. What you cannot buy inside the session is the part where somebody does the thing at their desk next week. A generic deck makes this worse in a specific way: because nothing touched the participant’s own tasks, no half-finished piece of real work is waiting on Monday to pull them back in. The transfer problem starts from zero instead of from something.
It is the same confusion that lets a license rollout get counted as a capability — the argument in the case against mistaking licenses for capability. Access and the ability to use it are different purchases, and a generic session is the second priced like the first.
The second invoice
Here is the cost no pitch mentions, including from firms selling the alternative.
A generic half-day for 25 people runs roughly $7,900 in payroll plus a fee. Call the fee $6,000 and the all-in number is near $14,000. Three weeks later nothing in anyone’s output looks different, so you buy the session you should have bought the first time: the same $7,900 of payroll, a second fee, and the first $14,000 written off. The cost of getting this wrong once is not the price of the bad session. It is roughly twice the budget you thought you were approving.
The harder part is not financial. You have already told the organization that AI enablement matters, booked the calendars and produced nothing visible, so the second request is a worse sell than the first. Whoever signed off now has a reason to say the training does not work, and they are holding a satisfaction report saying everyone liked it, which makes the failure impossible to locate. Budget for AI enablement is not an allocation that resets each year; it is credibility, and a generic session spends it without buying anything.
Three weeks is the right interval to check. If output has not changed by then, the diagnosis is almost always the pre-work, the task choice, or whether a single manager ever asked to see the work — the practical version being our look at what a good session actually has people do.
What a generic session costs in goodwill
There is a per-person cost that shows up in the next request, not this one.
A LinkedIn survey reported by Fortune in August 2025 found that 51% of professionals consider the intensity and frequency of AI training requirements excessive, with complaints centered on dense modules, tight deadlines and unclear practical payoff. In the same survey, 33% said they feel embarrassed by how little they understand and 35% said they are nervous discussing AI at work. LinkedIn published no sample size or field dates, so treat it as a survey of its own audience rather than a population estimate — but the direction matches what anyone who has stood in front of these rooms has seen.
Together those findings describe the failure mode of a generic deck. A third of the room arrives quietly worried they are behind. The session opens with how language models work and a tool tour, which teaches nothing usable and confirms their suspicion that the gap is theirs. They leave having been told AI is important, having done nothing, and having learned the next session will also be a presentation. The second invitation gets a lower acceptance rate, and that drop is the compounding cost.
A session that opens with the participant’s own recurring task inverts this, which is the design argument in our guide to building a session that changes Monday morning.
The full line item
A half-day generic session for 25 people, fully accounted:
| Line | Amount | Basis |
|---|---|---|
| Vendor fee | $2,000–15,000 | Market range for a hands-on session |
| Participant compensation, 4 hours | ~$7,900 | 100 person-hours at BLS $78.85 |
| Internal coordination | ~$550 | About 7 hours of one organizer’s time at the same rate |
| Licenses already paid for, still unused | Varies | Not caused by the session, but not fixed by it either |
| Follow-up | $0 | Rarely budgeted, which is why the 0.77 drop happens |
| Re-run, if nothing changes | Repeat of all of the above | Plus a harder internal approval |
The middle rows are the ones no proposal shows you. Once the payroll line is on the page the decision changes shape: when the room costs more than the trainer, the room’s own material is the only sensible thing to build the session from, and the comparison that matters is not fee against fee but fee against the chance of needing a second one.
Expect two to three hours of build per hour delivered when material runs on real client documents. A session quoted at the bottom of the market range with no pre-work request is not a bargain; the preparation you are not paying for is the preparation that would have made it work. Our timed 90-minute agenda shows what that preparation buys minute by minute.
When a catalog session is the right purchase
Public per-seat courses are the correct buy in two situations. The first is one curious person who wants orientation before anything is rolled out: Noble Desktop lists a six-hour Generative AI with ChatGPT course at $299 and a twelve-hour AI for Business with ChatGPT at $549; Webucator publishes a one-day generative-AI-for-business course at $779.10 per student, with a private on-site class offered separately. For one or two people that is the cheapest way to find out what the category contains.
The second is a leadership orientation that is explicitly not training. If five executives need shared vocabulary before approving a plan, a well-made generic presentation is the right artifact — label it an orientation and do not count it as enablement.
The arithmetic turns against the catalog around eight to ten participants. Twenty-five seats at $299 is $7,475 in fees alone, on top of the $7,900 in payroll, for content built for nobody in particular — custom-engagement money for catalog material.
Two more questions for the proposal
Beyond the pre-work test, ask these before signing and judge the answers rather than the slides.
What is the hands-on fraction, by minutes? Add up the minutes in which a facilitator or a slide is the thing moving, then the minutes in which every participant is typing against their own material. If the second number is not the larger, you are buying a presentation. A vendor who cannot produce the breakdown has not timed their own agenda.
What happens in the three weeks after? Not a portal, not recordings. A named person answering questions, and one manager who has agreed to ask to see the changed work. Neither is a training purchase, which is why no vendor volunteers them.
A vendor who can describe the pre-work, the minute breakdown and the follow-up is pricing preparation. A flat per-head rate with none of it is pricing a day of delivery from material that already exists — a reasonable business, and a poor fit for $7,900 of your team’s attention.
Frequently asked questions
What does a generic AI workshop actually cost?
The fee is usually the smaller half. A half-day for 25 managers carries about $7,900 in participant compensation at the BLS June 2026 rate of $78.85 per hour worked for management and professional occupations, against a market fee range of $2,000 to $15,000. Add coordination and re-run risk and the all-in figure is roughly double most approved budgets.
How can I tell whether a session is built from a generic deck before I book it?
Read the pre-work request. If participants are asked to bring nothing but a laptop, the material cannot be about their work. Then ask what changes in the room if they do bring something — a vendor with no answer has a form rather than a design.
What are the most common AI workshop pitfalls?
A session treated as the deliverable with no follow-up; tool demonstrations in place of the participant’s own task; a room too large to practice in; mixed skill levels with no streaming; satisfaction scores as the success measure; and content organized around how the technology works rather than what the team does on Tuesday.
Is a cheap catalog course ever the right choice?
Yes, for one or two people who need orientation, or for a leadership briefing labeled as a briefing. Published per-seat prices for this material run from roughly $299 for six hours to $779 for a day. An AI workshop for teams of eight to ten and up needs custom material; the per-seat math stops working.
How long should we wait before deciding a session failed?
About three weeks, on the tasks the session covered. If nothing in anyone’s output looks different by then, the problem is the pre-work, the task choice, or the absence of anyone asking to see the work.
Why do satisfaction scores stay high when nothing changes?
They measure the experience, not the behavior. The transfer research found an average within-study decrease of 0.77 from learning measures to behavioral measures across studies that tracked both. People learn the material and then do not use it, which a post-session survey cannot detect.
Does a longer session fix a generic one?
No. A full day of the same material for 25 people costs about $13,800 in payroll instead of $7,900 and changes the same amount. Length is not the variable; whose work is on the screen is.
What should the hands-on fraction be?
More than half the clock, measured in minutes of participants typing against their own material. Figures circulating as research findings here are mostly vendor assertions, so set the fraction from what each block is for and ask for the breakdown.
How much preparation should a custom session involve?
Two to three hours of build for every hour delivered, when the material runs on your documents. A quote at the bottom of the market range with no pre-work request does not include that preparation.
Key takeaways
- The fee is the smaller number. A half-day for 25 managers is about $7,900 in compensation at the BLS June 2026 rate of $78.85 per hour worked for management and professional occupations.
- Four hours spends close to 30% of a typical employee’s formal learning hours for the year, against ATD’s 13.7 hours used and $1,254 of direct spend per employee.
- “Generic” has a checkable definition: a session built on your work has to ask for your work. The pre-work request, not the agenda, is the tell.
- The loss happens after the room. Across 1,152 effect sizes from 165 sources, the within-study drop from learning to behavioral measures averaged 0.77.
- Getting it wrong once costs about twice the approved budget, and the credibility to ask again cannot be re-purchased.
- A catalog seat at $299 is right for one person and wrong for twenty-five; the arithmetic turns around eight to ten participants.
Before approving a quote, work out the payroll number for the room and set it beside the fee. A free AI-readiness call will tell you which team and which task the session should be built around, what the pre-work request needs to say, and whether a custom session or a handful of catalog seats is the cheaper way to find out.
Arthur Wandzel