Who authorises the people teaching AI?
There is currently no standard governing who is qualified to teach or certify AI use.
Anyone can issue a certificate. This article sets out the difference between capability and authority, why the distinction is a governance problem rather than a marketing one, and the three questions any AI credential should be able to answer.
The TL;DR version
Capability is what a person can do. Authority is what a person has been trusted to hold on behalf of others. AI education has collapsed the two, and the collapse is now producing real governance exposure for the organisations buying training.
Why is this a governance question and not a quality question?
Poor training is a quality problem. You waste a day and you know it by lunchtime.
Unauthorised certification is a governance problem, and it behaves differently. It moves downstream and it doesn't announce itself.
Here is the chain. A consultant completes an AI course and receives a certificate. The certificate carries no defined standard behind it, but it looks like every other certificate. The consultant is engaged by a school, a council, a health provider or a small firm. They deliver training. Staff go back to their desks and change how they handle client records, what they disclose, and what they let a model decide unsupervised.
The organisation now has an AI practice shaped by a person whose competence nobody verified, and no way of reconstructing why any of it was considered safe. If it goes wrong, the question asked will not be "was the trainer good". It will be "on what basis did you rely on them".
Most organisations cannot currently answer that. Not because they were careless, but because the market gave them nothing to check.
What is the difference between capability and authority?
Capability Authority What it is What you can do What you have been trusted to hold for other people How it's acquired Practice, over time Granted, deliberately, by someone accountable Who carries the risk You The people who act on what you told them How it's verified Demonstration of skill Demonstration against a defined standard What removes it Falling out of practice Failing the standard
Capability accumulates on its own. Authority never does. That is the whole distinction, and it is why "I've been using this for two years" is a true and completely irrelevant answer to the question of who authorised you to teach it.
What should an AI credential be able to answer?
Three questions. Ask them of any AI credential, including ours.
1. What did the holder have to demonstrate to earn this?
Attendance is a receipt. Completion is a receipt. Demonstration in front of an assessor is a credential. If the answer is that they turned up and paid, you are looking at proof of purchase.
2. Who assessed them, and against what?
A standard written by the person selling the course, existing nowhere outside the sales page, is a description rather than a standard. Ask to see it. Ask who else is held to it.
3. What happens if the holder teaches it badly?
Every established profession has an answer. Most AI certifications have never been asked. If there is no route to review, revoke, or even record a complaint, then nothing was ever really granted.
A credential that cannot answer these may still sit behind a genuinely good course. It simply isn't authority, and it should not be presented as though it is.
Doesn't a standard just keep people out?
It can, and this is the failure mode worth guarding against as carefully as the first one.
A standard set too high becomes gatekeeping with better manners. Prerequisites, panels, years-of-experience floors and fees that quietly sort by who already had capital. The predictable outcome is that AI education gets taught by the people who were already in the room, which reproduces exactly the concentration of authority that ethical AI work exists to interrupt.
The design principle is that the standard is a floor rather than a wall. Set honestly, applied identically, and visible from where an applicant is currently standing so they can see what crossing it requires.
It is also worth saying plainly that time in AI is a poor proxy for readiness to teach it. Judgement, ethics, instructional skill and knowing when to stop are the load-bearing capabilities, and they are frequently strongest in people who came to AI recently from a discipline where the stakes were already real.
What this means for an organisation buying AI training
A short due diligence pass, which takes about ten minutes.
Ask the trainer what they were assessed against, and ask to see it
Ask who holds them to it, and whether that body is independent of the sale
Ask what happens if the training turns out to be wrong
Record the answers alongside the engagement, so the basis for the decision can be reconstructed later
Treat "certified" with no named standard behind it the same way you would treat any other unverified supplier claim
None of that requires a legal team. It requires asking the questions once and writing down what came back.
How SHE IS AI approaches this
The Applied Intelligence Pathway™ places people by demonstrated readiness rather than by what they choose to pay for. There is an assessment and an interview before placement, which is slower and more costly to operate than a checkout page, and is the reason the credential means anything.
Catalyst, the first level, opens on 15 September. Six weeks, twelve modules, live.
