What an AI prompting course costs and how long it takes

On-Off Group

September 2026

AI prompting course cost is driven by four things: how long the session runs, how many people are in the room, whether the content uses your own work, and whether anyone follows up afterwards. Most prompting courses run from two hours to two days.

If you are pricing up AI prompting training for your team, here is what drives the cost, how long it runs, what your people need to bring, and how you show your boss it did anything.

Search for ai prompting course cost and you get figures with no context. A two hour session and a two day programme sit on the same results page. So does self-paced video. None of it tells you what you will be paying for.

If you have to book the training and then answer for it later, the four cost parts below are the ones to check.

The four things that set the price

Almost every quote you get for prompting training is built from the same four parts.

Length. Two hours, half a day, a full day, or a day plus a follow-up session a few weeks later.

Group size. If you want everyone typing during the session, the room has to stay small enough for the facilitator to get round. A big room turns into a talk.

Custom content. Generic prompting exercises are cheap because they already exist. Building the exercises from your own documents, your own customer emails, your own underwriting notes, takes preparation time and costs more.

Follow-up. A session with nothing after it is the cheapest option on the list and the one most likely to be forgotten. Our page on custom workshop pricing, length and what you need to prepare sets out how those four parts stack up when the session is built for your team.

When you compare two quotes, compare them part by part. Otherwise one quote is a day of practice and the other is a recorded video, and the prices look the same.

How long it actually takes

Two hours is enough for one task. People learn to write a clear instruction, give the model an example, and check what comes back. That is a real skill and two hours is enough for it.

A full day is worth it when at least half of the time is spent on the team's own work. If the day is mostly demonstrations, you have bought a longer version of a video.

Self-paced courses are the cheapest per head and the slowest to show up in the work, because nobody is sitting next to you when the output is wrong.

The part people underestimate is the gap after. People pick up prompting on real work after the session, not in the room. A short session a few weeks later, where people bring what broke, does more than adding hours on the day. Budget for that rather than for a longer single session.

What your team needs to bring

This is where sessions get eaten up by admin.

Access. Accounts created, logins tested, approvals cleared before the day. If the AI tool is blocked on the corporate network, you want to know that before people are sitting in the room.

One real task each, not a general topic. A claims note to summarise. A branch memo to rewrite. A set of customer complaints to sort into groups. People who arrive with nothing spend the session watching.

Permission on data. Say plainly which documents can go into the tool and which cannot. Put it in the slides at the start so nobody has to guess.

A baseline number for at least one of those tasks. More on that next.

We ask for the same four things when we build an AI feature with a client team, and for the same reason: what building an AI feature with us involves and how long it takes starts with access and real examples too, because without them the first day gets spent on admin.

Write down a before number

Most teams cannot tell you whether their last training worked. The room was full and the feedback scores were good, but that's it...

So pick one task and measure it before the session. How long does it take now, end to end? How many go back for rework. How many cases a person clears in a day. Turnaround time on an onboarding case is a good one because your boss already asks about it.

Then measure the same thing four to six weeks later, on the same task, counted the same way.

Small honest numbers hold up. On a series of five day design sprints with a large insurer, customer journeys that used to take six months or more were designed, tested with customers and live in four weeks, and completion rates on the new journeys rose 80 per cent, measured by the client. Those figures only exist because somebody wrote down the old ones first.

If the number barely moves, say so. That is cheaper to find out now than in six months.

Why the cheap session often costs more

A cheap session for a big room that changes nothing costs you more than a paid day for a smaller group who use it, because you still have to show your boss a result.

The test is the first real piece of work after the room empties. Did anyone open the AI tool on that task, and did it help? The certificate tells you nothing.

Two things move that. First, people practise on their own work in the room, so the prompt they wrote is useful on Monday morning, not a sample exercise about a made up company. Second, someone owns the follow-up. A named person, a date in the calendar, and one task each that gets checked. Without that, the good session and the bad session look identical after a month.

We have run workshops for Philippine banks and insurers since 2015. The sessions that carry on into the work are the ones where a team lead arrived with a task list and a number.

What to do this week

Pick one task your team does every day and write down how long it takes now.
Check whether the AI tool your team would use is actually approved and reachable on your network.
Ask each person to bring one real piece of their own work to the session.
When you compare quotes, compare length, group size, custom content and follow-up separately.
Put a date in the calendar four to six weeks after the session to measure the same task again.

Questions people ask

How long should an AI prompting course be?
Two hours is enough to get people writing better prompts on one task. A full day only pays off if half of it is spent on the team's own work rather than on demonstrations.

What makes one AI prompting course cost more than another?
Session length, group size, whether the exercises are built from your own documents and tasks, and whether there is any follow-up after the room empties. Custom material and follow-up cost more and are usually the parts that change Monday.

What do we need to prepare before the session?
Accounts and access sorted in advance, one real task per person, and a before number for that task such as current turnaround time on an onboarding case.

How do we know the training worked?
Check the first real piece of work after the session. Did people open the AI tool on that task, and did the number you recorded beforehand move? Feedback scores only tell you people showed up.

On-Off Group trains teams, tests products with real customers, finds where a transformation has stalled and builds what gets it moving, for banks, insurers and enterprises in the Philippines, since 2015. How we help with ai prompting.