AI and CPD: What Actually Changes for Training Providers

Artificial intelligence has changed how training is produced faster than it has changed how training is recognised. Courses can now be drafted in an afternoon, and certificates have never been easier to generate or to fake. The effect on CPD is not that credentials matter less. It is that the credential's ability to prove something specific, and to be checked by whoever receives it, has become the part that carries the weight.

What AI genuinely changes for training providers

The useful applications are unglamorous and real. Drafting course structure from material you already have. Turning a rambling set of session notes into clear learning outcomes. Producing variants of a course for different audiences without rewriting it from scratch. Handling the administrative sludge around scheduling, reminders and follow-up.

These are time savings on the parts of the job that were never the point. A trainer who spends less time formatting a workbook has more time to teach, which is a straightforwardly good outcome.

Open CPD's own use of this is deliberately narrow. CPD Genie reads training material a provider already has and proposes the aims, skills and outcomes that describe it. It does not write the training and it does not judge it. It helps with the part providers most often say they find hardest, which is putting into precise words what their own course does.

What AI does not change

It does not make a course good. A well-structured document about a subject is not the same as a well-taught session, and no model has the experience of a practitioner who has watched two hundred people get the same thing wrong in the same place.

It also does not confer authority. A course generated quickly still needs someone willing to put their name to it and stand behind what it claims to teach. That signature is the thing being recognised, and it cannot be automated because its whole value is that a real organisation is accountable for it.

The recognition problem AI has sharpened

If a plausible-looking course and a plausible-looking certificate can both be produced in minutes, then looking plausible stops being evidence of anything. That is the practical consequence for CPD, and it cuts both ways.

For providers doing careful work, it means the burden of proof has shifted onto the credential. A certificate that names a course and a date now says roughly what an unsourced claim says. A certificate that states what the training aimed to do, what skills it built, what the learner can demonstrate, how many CPD hours it represents and who is accountable for all of that, and which resolves to a record the recipient can open and check without asking anyone's permission, says something an AI-generated file cannot imitate. The difference is not the design. It is that one of them is connected to a live record and the other is a picture.

For learners and employers it means checking is no longer paranoid. Verifying a credential in under a minute is now a reasonable default rather than an insult, which is a healthy shift for everyone issuing honest credentials.

Using AI in your training without weakening the credential

• Keep a person accountable for content. Use AI to draft and restructure, then have someone who knows the subject read every line before it reaches a learner. The declaration you sign as a provider is about legitimate authority over what you certify, and that is a human obligation.
• Be specific in your outcomes. Generic outcomes are the easiest thing for a model to produce and the least useful thing for an employer to read. Specificity is where the value has moved.
• Say what changed and when. If AI helped you update a course, update the aims, skills and outcomes with it. A credential describes a version of a course, not a course in perpetuity.
• Do not let efficiency become volume for its own sake. Issuing more certificates faster is only an improvement if each one still corresponds to something a learner actually did.

Where this leaves CPD

The optimistic reading, and the accurate one, is that AI has raised the value of anything that can be verified. The credentials most exposed are the ones that were only ever a design and a date. The credentials that gain are the ones carrying specific claims backed by a checkable record, because those are now doing work that nothing else can do cheaply.

For a provider, the response is not to avoid AI. It is to use it on the drafting and keep the judgement, the accountability and the evidence firmly attached to your own name. How it works sets out the structure, and the verification page shows what the checking looks like from the other side.