A training certificate is not enough proof that the training worked
Articles › Training evaluation
A certificate only shows someone attended. The proof is the first real piece of work after the room empties: did they open the AI tool for it, and did it help?
Book AI training this year and you will get certificates back. A training certificate is not enough to tell your boss whether anything changed. It tells you someone sat in the room for two days and passed whatever the room asked of them.
This is for the people who have to answer the next question. Did the work get faster? Did anyone use the tool on a real task. Is the turnaround time on your onboarding flow any different from the week before?
Most of this has to be set up before the training runs, not after.
What this looks like on a normal working day
The training runs for two days. The room is full, the feedback scores are good, the photos are good, and the certificates go out the week after.
Then Monday. Everyone opens the same files they had open before. The prompts from the exercise are in a handout nobody reopens. Three weeks later someone asks how the AI training went and the honest answer is that people enjoyed it.
We have run workshops for Philippine banks and insurers since 2015, and this is the most common pattern. The training itself is usually fine. It just has nothing hooked into real work. The room is a nice place to be for two days and the work outside the room is unchanged, because nobody agreed in advance which piece of work it was supposed to change.
A better course will not fix that. Decide before anyone books a room which piece of real work the training is supposed to change.
Pick the first real task before you book the training
Before the training, name one task that someone will actually do the following week using what they learned. A task, not a topic. Drafting the customer email replies for the disputed transactions queue. Writing the first pass of test cases for the new pilot. Summarising branch staff feedback from the last quarter.
Write the name of the person next to it. Write the week they will do it.
This sounds obvious and it is almost never done. Requests usually arrive as a headcount and a topic: forty people, AI prompting, two days. When we ask which piece of work should be different afterwards, people usually have to go and find out. That conversation is worth having on its own. If nobody can name a task, that is worth knowing before the purchase order goes out.
When the task is named, the training changes shape. The exercises use that task. People leave with something half finished that they want to go back to, which is a better outcome than a full set of notes.
Get a before number, even a rough one
Most teams we work with have no baseline. Nobody asked for one until after the project ended, and by then the old numbers are gone.
So take the number before the training. Rough is fine. How long does that task take now, across five recent examples? How many go back for rework. How many days between a customer submitting and getting a decision?
Five examples and a stopwatch is enough. You are not publishing research. You are giving yourself something to compare against in six weeks.
Here is a before and after we can stand behind. A large insurer ran five day design sprints on customer facing web journeys. Journeys that used to take six months or more were designed, tested with customers and live in four weeks. Time to market fell from 26 weeks to 4. Completion rates on the new journeys rose 80 per cent, and the drop-off points that existed before were gone.
That was sprint work, not AI training, but the same rule holds: those numbers exist because someone wrote down the old ones. Without the 26 weeks, the 4 weeks is just a claim.
Check on Monday, not at the end of the quarter
Set a date about a week after the training and check two things. Did that person start that task? Did they open the AI tool while doing it?
If the answer is no, find out why in plain terms. Often it is access. The tool is blocked, or the licence is with IT, or the data they need is in a system they cannot export from. Usually a quick fix, but it never gets made because nobody looked until the quarterly review.
Sometimes the answer is that the task moved or the person got pulled onto something urgent. Fine. Pick the next task and the next week.
The point of checking early is that the training is still fresh enough to be rescued. A month later people have forgotten the prompts they wrote and all you have left is the certificate.
Open the sign-up instead of filling a register
Training that people are made to attend does not change how they work the following week. They attend, they are pleasant about it, they go back to what they were doing.
One test costs nothing. Open the sign-up to people who want it and see who puts their hand up. If the room fills, most of those people already have work they want to use it on. If it does not fill, the topic is not the problem people have.
Sun Life workshop participants from 2015 are now senior across the industry. They kept using it on their own work, years after the room emptied. That is a slow signal and no help to you this quarter, but that's what it looks like when training sticks.
What to do this week
- Name the first real task each person will do after the training, with their name and the week next to it.
- Take a before number on that task. Five recent examples, timed, written down somewhere you will find again.
- Ask whoever runs the training to build the exercises around that task, not generic examples.
- Put a check in the diary for one week after the session: did the task start, did the tool get opened, what got in the way.
- Take the after number six weeks later and report both, even if the gap is small. A small honest number beats a feedback score. More of the workshop scripts, and the projects they came from, are at onoffgroup.com.
Everything on Training evaluation
Questions people ask
So what should we measure instead?
Measure the first real task people use AI on after the training, and whether it actually helped. Take a number before the training so you have something to compare against later.
Why are attendance and feedback scores not enough?
A full room and a good feedback score tell you people showed up, but that's it. Neither says anything about how people work the following Monday.
What if people were told to attend?
Training people are made to attend does not change how they work on Monday. That is a big part of why certificates prove so little.
Where can we see more of this thinking?
More workshop scripts, and the projects they came from, are at onoffgroup.com.
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 training evaluation.
