IAPP's AIGP certification has gone from niche curiosity to a fixture in Singapore and Malaysia corporate training conversations over the past year or so. I teach this material, so take my enthusiasm with the appropriate discount, but the momentum is mostly deserved. The content is solid. I've taken enough compliance professionals through the EU AI Act's risk tiers and the provider-versus-deployer obligations to know that most of them genuinely didn't have this vocabulary before, and these are people whose organisations were already deploying systems the Act would classify. Giving them a rigorous shared framework is worth doing. No argument there.
The problem I want to name is smaller and more human, and I noticed it the uncomfortable way. About a year ago I had a cohort where, roughly an hour in, it dawned on me that half the room was there because a manager had told them to be. No live AI project, no vendor decision pending, no system anyone would ask them to assess. They were perfectly pleasant. They took notes. Some of them passed the exam. And I'd bet a reasonable sum that within three months, very little of it survived contact with their actual jobs, because none of it had anything to attach to.
I want to be fair to those participants, because my first reaction, mild irritation at a disengaged room, was the wrong one, and it embarrasses me slightly in hindsight. Adult learning without an application context decays fast. That's not a character flaw; it's just how memory works. I've personally forgotten entire frameworks I learned without a live reason, some of them expensively. The disengagement wasn't their failure. It was a purchasing decision made two levels above them, buying a credential for a board deck rather than a capability for a decision.
Because that's the contrast that makes the pattern obvious. The participants who arrive holding a real question, we're procuring an AI screening tool next quarter, is it high-risk under the Act? what do we owe as a deployer versus what the vendor owes as a provider? those learners argue with the material, stress-test it against their case in front of everyone, and leave with something that works. The conscripted cohort leaves with a line for someone's LinkedIn profile. Same syllabus, same instructor, same coffee. The variable was never the content.
Exhibit · Same syllabus, same instructor
What separates who it sticks to
Sent without a live case
A credential for the board deck
- Perfectly pleasant, takes notes
- Some pass the exam
- Little survives contact with the job within three months
Sent with a live case
A capability for a decision
- A pending vendor decision or system to classify
- Argues with the material, stress-tests it
- Leaves with something that works
There's a longer-term worry too, and it's self-interested, so I'll declare it. Credentials hold value because the market believes holders can do something. If a wave of certification is driven mainly by board-deck optics, the signal degrades for everyone, including the professionals who earned it with genuine intent, and including trainers like me whose work rides on the credential meaning something. We've watched this happen to other compliance certifications. It would be a shame to rerun the film with AI governance, at precisely the moment the underlying skill starts to matter.
My advice to anyone budgeting for AI governance training fits in a sentence: send people with a live case. A pending deployment, a vendor evaluation, a system inventory that needs classifying, anything real the framework can be tested against during the programme and in the weeks after. And if no such case exists yet, that's worth sitting with honestly, because it suggests the certification is answering a communications need rather than a governance one. The training will still be here when there's something for it to govern.
If you take one thing: certification sticks to a live decision and slides off everything else. If you're sending people without one, ask what you're actually buying.
