What AI training for a team costs and how long it takes

On-Off Group

September 2026

AI training cost in the Philippines depends on three things: how many people are in the room, how many days you run, and whether the exercises use your own work or generic ones. Most team programmes run half a day to three days, plus a week of prep.

If you have been asked to put a number in a budget line for AI training, this is for you. It covers what drives the price, how long the whole thing takes, and what you need ready before anyone quotes you.,

No prices here. Quotes vary too much by head count and format for a figure in an article to be worth anything to you. What you can take from this is the shape of the spend, so the number you put in front of your boss holds up when questioned.

On-Off Group has run workshops for Philippine banks and insurers since 2015, so most of this comes from rooms in Makati and BGC rather than from a pricing survey.

Three things that move the price: head count, days, and how custom it is

When a quote comes back higher than expected, it is almost always one of three things.

Head count. A room of 15 to 20 people is one facilitator. Above that you are paying for a second one, or you are paying for the same session run twice. Sixty people is three sessions, not one big one, if you want anyone touching a keyboard.

Days. Half a day gets people comfortable enough to start typing prompts. Two to three days gets a team producing work they would actually send. The gap between those two is the biggest single lever on cost.

How custom it is. A standard course costs less because the exercises already exist. If you want people practising on your onboarding flow, your claims notes, your branch staff scripts, someone has to sit with your material first and build the exercises. That prep is real time and it shows up in the quote. It is also the part people still use weeks later. There is more on custom workshops, length and what to prepare at onoffgroup.com.

The calendar is longer than the training days

People budget for the days in the room and forget the two weeks around them.

Before: roughly a week to collect the real tasks. That means asking each team what they actually spend Monday morning doing, and picking one task per team to build the exercises around. It is short calls and a shared document, not a project.

The room: half a day to three days, depending on what you bought.

After: four to six weeks before you can tell whether people are using it on real work. That is when people either use it on live work or quietly stop. Someone needs to check in, look at what people produced, and answer the questions that only turn up once people try it on live work.

So a two day programme is realistically a six to eight week line in your plan. If a proposal shows only the training days, ask who is doing the before and after, because if the answer is nobody, you will be back in the same room next year. What corporate AI training covers gives the range of formats if you are still deciding which one to price.

Get a before number or you cannot tell your boss whether it worked

Cost is usually not what goes wrong. It is that six weeks later, when the boss asks whether it worked, the only evidence is an attendance sheet and a feedback score. Those tell you people showed up, but that's it.

Pick one task per team. How long does it take now. How many go back for rework. What is the current turnaround time on the thing you are trying to speed up. Write the number down before anyone walks into the room, even if it is rough and taken from five people rather than a system report.

Then measure the same task after. If it moved a little, say so. A small honest number will hold up when your boss questions it. A vague claim will not.

This is also the cheapest part of the whole exercise. It costs a few conversations. Skipping it is what turns a defensible budget line into a slide pack.

What you need ready before the training starts

Three things, and the training is weaker without any of them.

Access. People need to be able to open an AI tool on their work machine on the day. If IT has to approve it, start that before you book. A room full of people watching a demo because the tool is blocked is a wasted budget line.

Real work. Bring actual documents, actual customer complaints, an actual process people hate. Exercises on made-up material do not carry over to the real work.

A reason each person is there. Training people are made to attend does not change how they work. If someone cannot say what task they are hoping to do faster, they are the wrong person for that seat. Buying licences for everyone and hoping people use them fails the same way. There is a separate article on that at onoffgroup.com.

When training is not the thing you need

Sometimes the honest answer is that a course will not fix it.

If the problem is that one process is slow and you already know which one, building a rough working version of an AI feature for that process may get you further than training thirty people. You pay for fewer people and more build time. What building an AI feature involves and how long it takes covers that side.

If the underlying customer journey is broken, AI will make the failure happen faster. Fix the journey first, then train people on the version that works.

Train people when you want a lot of them doing small tasks faster, and you can name those tasks.

What to do this week

Ask three teams to name the one task they would most like to do faster, and write the current turnaround time next to each.
Decide your head count per session. Cap it at around 20 if you want people building rather than watching.
Check now whether an AI tool is approved and open on staff machines, and start the IT request if it is not.
Put six to eight weeks in the plan, not two days, and name the person who will check what people produced four weeks later.
Price two versions, a short standard one and a custom one built on your own material, so the budget conversation has a choice in it.

Questions people ask

How long does AI training take?
Most team sessions run from half a day to three days of room time. Add about a week beforehand to collect the real tasks people will practise on, and a check four to six weeks later to see what actually got used.

What makes one quote higher than another?
Head count, number of days, and how much of the material is built around your own work. Custom exercises using your onboarding flow or claims process cost more to prepare than a standard course, and they are the part people still use on Monday.

What do we need to prepare before booking?
A named task per team, a rough before number for that task such as current turnaround time, and confirmation that people can access an AI tool on their work machines on the day.

How do we know if the training worked?
Pick the first real task people will use AI on, write down how long it takes now, then measure the same task four to six weeks later. Attendance and feedback scores tell you people showed up, not that anything changed.

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 training.