Book the half day first. For most teams the shorter session teaches as much as the longer one, and the money you were about to spend on hours five through seven is worth far more held back for a second session three weeks later.
That is not what the market will tell you, because the market sells days. Of the four training providers whose current pages we checked, three price a seven-hour unit as their default and none publishes a half-day rate. A buyer typing “half day AI workshop” into Google is shopping for a unit the catalog barely stocks.
The answer the vendor menu will not give you
Half-day and full-day sit side by side on most service pages as two tiers of one product, where more hours means more value and a bigger invoice. Nobody selling both has a reason to tell you which to buy first. Here is what the published pages say, checked this month:
| Provider | Published length | Published price |
|---|---|---|
| BrainStation AI Fundamentals | Full day in person, 9:00 to 4:30 | Not listed; team pricing by request |
| Certstaffix, Making ChatGPT and Generative AI Work for You | 1 day | $350 per person, group price on quote |
| NobleProg AI Overview | 7 hours | $1,344 online, $1,944 classroom, per participant |
| Section AI Crash Course | 1.5 hours live, spread over one to two weeks | $195 list |
The seven-hour day is the unit of account across the corporate AI workshop market, and the one provider that sells short does not sell a half day either. It sells ninety minutes of live instruction stretched across a one to two week sprint, a different design decision with better research behind it.
Private group rates are quoted rather than published, so the half-day and full-day spread is rarely visible before a call. Market rates for a facilitated private session run $2,000 to $15,000 depending on how many roles need their own material, and our breakdown of what a one-day session costs walks the line items.
What a half day fits, and what a full day fits
Strip the marketing and the two formats differ by about three contact hours. What those hours hold is more specific than any brochure admits.
| Half day | Full day | |
|---|---|---|
| Contact hours, net of breaks | 3 to 3.5 | 5.5 to 6.5 |
| Workflows a group can practice well | 1 to 2 | 3 to 4 |
| Realistic group size per facilitator | 12 to 25 | 12 to 25 |
| Lecture to practice ratio that holds | 1:2 | 1:2 before lunch, worse after |
| What people walk out holding | Two tested prompts for one real task | A small library, unevenly tested |
| Calendar cost per attendee | A morning | A working day plus the commute |
| Chance of a second session being funded | High | Low, the budget went on day one |
A half day teaches one workflow to the point where people can repeat it on Monday. Pick the task the team does most and hates most, often turning raw notes into a client-ready summary, and spend the morning on that one thing with their own documents open.
A full day covers three or four workflows, a real gain when the room holds distinct jobs. The cost is that the last two hours are the weakest teaching of the day and the most expensive, because the room’s attention budget is already spent.
You are buying the room, not the day
Booking an AI workshop for teams is a labor decision in procurement clothing. Every vendor page prices the facilitator; none prices the people in the chairs, which is where the real money is.
Twenty-five attendees in a seven-hour session is 175 person-hours. At a loaded $60 an hour that is about $10,500 of payroll sitting in the room, which on most engagements is the larger half of the bill and the half nobody invoices. The same group in a 3.5-hour morning is 87.5 person-hours, roughly $5,250. The saved time covers a second session at market rates with change left over.
That reframes the question. It is not whether you want more content for more money. It is whether hour six is worth another $1,500 of your own payroll on top of the fee. The answer is usually no, and the saved hours are better spent on a session where the team arrives with real questions.
LinkedIn’s 2025 Workplace Learning Report found employees averaging 40 hours of training across a year, down from 47. A full day spends roughly a sixth of a person’s annual training time in one sitting. ATD’s 2026 State of the Industry report put direct learning spend at $846 per employee across 340 organizations, so the fee side of this call is small and the time side is not.
The evidence says split the hours
The strongest argument for the half day has nothing to do with budget. Hold contact hours constant, split them across two sittings, and people retain more. It is one of the oldest and best-supported findings in learning research.
Cepeda, Pashler, Vul, Wixted and Rohrer synthesized 839 assessments across 317 experiments in 184 articles in Psychological Bulletin in 2006. Spaced study episodes beat packed ones, and the gap that produces peak retention widens as the interval before the test widens. The longer you need a skill to survive, the wider the gap between sessions should be. If you want the team still using a workflow in six months, two or three weeks apart is closer to right than three hours after lunch.
A 2025 meta-analysis in Behavioral Sciences reached the same place in real classrooms rather than labs, reporting an effect of about d = 0.54 for distributed over massed practice across 22 classroom studies and more than 3,000 students. That is a large gain for a change that costs nothing but calendar work.
Two mornings three weeks apart buy the same contact hours as one full day. Same payroll, close to the same fee, better retention. Vendors sell the single day anyway because one booking is easier to close than two.
What happens in week three is the real decision
Saks and Belcourt surveyed training professionals across 150 organizations and reported that 62% of trainees applied the material right away, 44% at six months, and 34% a year later. That decay is what you are fighting, and against it the length of one session is a second-order variable while the week-three touchpoint is the first.
You will also hear that only 10% of training ever transfers to the job. Treat that number with suspicion. It traces back to a 1982 article that offered no evidence for it and it has been repeated as fact ever since. The honest version of the claim is the Saks and Belcourt curve: most people do apply what they learned at first, and then it slides.
That slide is where a second session earns its money. People come back with the thing that broke, the prompt that worked twice and failed the third time, the colleague who refuses to try. None of those questions exist on day one, so a full day cannot answer them; it can only cover more untested ground. That gap between buying training and building real AI enablement is the whole argument of our guide to why license spend is not the same as team capability.
Which to book first, by team shape
The question answers itself once you know what is in the room.
Book the half day when:
- One team with one job needs one workflow. A finance group that wants faster variance commentary needs a morning, not a day.
- Nobody has a baseline yet. A shorter first session shows you where the real gaps are before you commit a bigger budget.
- The group is 12 to 25 people in one function.
- Attendance is the risk. A morning gets accepted by calendars that reject a full day, and a half-full full day is worse than a full half day.
- You intend to run a second session, so the hours stay in reserve.
Book the full day when:
- The room is cross-functional and the point is a shared artifact: agreed rules for safe use, a ranked list of workflows to automate, a draft standard the group argues its way to. That output only exists after the room has worked through disagreement, which takes longer than a morning.
- Travel concentrates the cost. If people are flying in, the marginal hour is cheap and the marginal trip is not.
- The team is already fluent and wants depth on chained workflows rather than a start.
- A second date is truly unbookable. One day beats a half day that never gets a sequel.
For leadership groups the scarce resource is calendar rather than money, so the format that works is usually shorter and more frequent; our comparison of multi-day immersion programs for leadership teams covers where the longer formats pay off. Whether the room is physical or remote moves the ceiling on both, which we take apart in on-site versus remote delivery.
What a room does at hour six
We run these sessions, so this part is observation rather than theory. A workshop has a shape, and it is the same shape almost every time.
The first hour is slow while people log in and stop performing for their manager. Hours two and three are the best teaching of the day: the room is awake, the examples are their own documents, and you can watch people get it. Lunch resets attention downward.
Hour five still works if the afternoon is pure practice with a facilitator walking the room. Hour six is where it goes. Slack comes out, the strong participants get bored, the quiet ones stop asking, and the facilitator talks more because the energy to practice is gone.
So a full day is three excellent hours, two decent ones, and one you paid full price for. A half day is the three excellent hours with the weak ones cut. That is why the shorter session so often produces the same Monday-morning change, and the design mechanics behind it are the subject of our workshop design playbook.
How to brief whichever one you book
Length matters less than the brief. The same two sessions land differently depending on what the facilitator gets in advance.
Send real material: three or four documents the team works with, redacted if needed. A client email thread, an intake form, a board summary, a messy spreadsheet. A session built on your documents teaches a workflow; one built on generic samples teaches a demo.
Name one outcome in plain language, the thing people should be able to do on Monday that they cannot do on Friday. If the vendor cannot restate it back to you, the agenda is theirs rather than yours.
Cap the room at 25 per facilitator. Hands-on AI training lives or dies on practice time per person, and past that number nobody gets corrected, which is the whole product.
Book the follow-up before the first session runs: three weeks out, 90 minutes, same group, diarized while the budget is open. It never gets booked later, and it decides whether any of this survives to month six.
Five questions to ask before you sign
- How many contact hours, net of breaks and introductions? Seven-hour days often hold five hours of teaching. Compare the net figures.
- What share of the time is hands-on practice? Under two-thirds and you have bought a presentation, at either length.
- Will you build the exercises from our documents? If the answer is a standard curriculum, length does not matter, because nothing will transfer.
- What does the half-day version drop? A vendor who cannot answer precisely has not designed two sessions, only priced one twice.
- What does a follow-up include and cost? The answer tells you whether you are buying attendance or capability.
Frequently asked questions
Is a half-day AI workshop long enough to be useful?
Yes, for one workflow with one team. Three to 3.5 net contact hours is enough to introduce a task, practice it on the team’s own documents, correct people one by one, and send everyone out with two prompts that work. It is not enough for three or four different jobs, which is the case for the longer format.
What does a half-day AI workshop cost compared with a full day?
Private group rates are quoted rather than published, and the market range runs about $2,000 to $15,000 depending on how many roles need their own material. In the quotes we see, a half day lands at 55% to 70% of the full day, because design and travel do not halve when the hours do. Per-seat catalog pricing is more visible: public one-day classes run from $350 a head to just under $2,000.
Should we run one full day or two half days?
Two half days, if you can get both dates in the calendar. Same contact hours, same payroll, and the spacing research favors the split. The 2006 synthesis in Psychological Bulletin found the best gap widening as the retention window lengthens, so two or three weeks suits a skill you want people still using next quarter.
How many people can attend a half-day session?
Cap it at 25 per facilitator, and 12 to 18 is better. The limit is practice time rather than room size: once the facilitator cannot reach everyone during the hands-on blocks, people stop being corrected and the session becomes a lecture with laptops open.
Do executives need a different format?
Shorter and more frequent, because their constraint is calendar rather than budget. A 90-minute session on the two decisions they personally own beats a full day of general content they will leave halfway through. Longer formats work for leadership only when the output is a decision the group must reach together.
What should the team walk out with?
From a half day: one workflow they can repeat, two or three tested prompts saved somewhere shared, and a named next task for that week. From a full day: the same across three or four workflows, with the caveat that the later ones get less practice. Slides are not a deliverable.
Do we need a follow-up session?
It is the highest-value hour you will buy. Application falls from 62% right away to 44% at six months in the Saks and Belcourt data, and the questions behind that slide only appear once people have tried the work alone.
Can we just buy a one-day course per seat instead?
For two or three people, yes, and it is cheaper. For a team it is the wrong shape: catalog classes teach generic examples, so each attendee has to translate the lesson back to your work alone. A private session practices on your documents, which is why it transfers.
Key takeaways
- Book the half day first. Three to 3.5 net contact hours covers one workflow properly, and the held-back budget buys a follow-up worth more than hours five through seven.
- Price the room, not the day. Twenty-five people for seven hours is 175 person-hours, about $10,500 of loaded payroll, which on most engagements outweighs the fee nobody argues about.
- Two half days beat one full day at identical contact hours. The 2006 Psychological Bulletin synthesis of 317 experiments and a 2025 classroom meta-analysis at about d = 0.54 point the same way.
- The week-three touchpoint decides whether anything survives. Application slides from 62% to 44% within six months, and the full day spends the budget that would have paid for the fix.
- Full day earns its cost when a cross-functional room has to produce a shared artifact, when travel concentrates the cost, or when a second date is truly unbookable.
Count the people who will sit in the room, multiply by the hours and a loaded rate, then compare that figure with the fee you were about to argue over. Then book the shorter session and diarize the second one. A free AI-readiness call will tell you which workflow belongs in the first morning, the only part of this decision that changes what happens on Monday.
Arthur Wandzel
