I hold a stack of AI certificates. Here is what they're worth.

I hold a stack of AI certificates and certifications.

That sounds more impressive than it is.

I am not against credentials. I took the work seriously. I completed the material. I passed what needed to be passed. That includes formal vendor exams and Google AI certificates covering fundamentals, brainstorming and planning, research and insights, writing and communicating, content creation, data analysis, and app building.

Some of it was useful. Some of it forced me to read product details I might otherwise have skimmed. Some of it gave me vocabulary for explaining AI to business owners who do not live inside this world.

But if you are choosing someone to help your business with AI, you should not stop at the certificate.

A certificate can tell you that someone completed training material or passed an exam. It can tell you they cared enough to sit through the material. It can tell you they know the vendor's language. Depending on the credential, it may also tell you they passed a proctored test under time pressure.

It does not tell you they can protect your client data.

It does not tell you they can review AI-generated code before it touches a live system.

It does not tell you they will talk you out of a bad idea.

It does not tell you they understand your business.

What certificates are good for

A certificate is a receipt.

That is not nothing. Receipts matter. They show that a person did the work at least once. They show a baseline of interest. They may show familiarity with products, terminology, controls, and common use cases. In a field full of people making loud claims after a weekend of experimentation, a receipt has some value.

Vendor certifications can also be useful because they expose how the vendor wants the tool understood. Google Cloud's Generative AI Leader exam, for example, is a proctored multiple-choice exam with a defined topic list, time limit, and validity period. Microsoft's AI fundamentals exam is designed for people with technical and non-technical backgrounds. Anthropic's Claude Certified Associate, Foundations, is a formal exam around practical Claude use, evaluation, product selection, workflows, and responsible AI principles.

Those details matter. They are more meaningful than a badge that simply says someone watched videos.

But even a serious foundational exam is still foundational. It proves contact with the material. It does not prove the person can operate under messy conditions.

What certificates miss

Real AI work does not happen inside a clean exam question.

It happens when a business owner has a messy spreadsheet, unclear process, nervous staff, client information in the wrong places, and no technical department to review the output.

It happens when the AI gives code that looks right but references sheets and ranges that do not exist.

It happens when a draft policy sounds professional but quietly allows employees to paste confidential information into personal accounts.

It happens when a tool can do the task, but the business cannot maintain the workflow after the consultant leaves.

It happens when the right answer is "do not build this yet."

Certificates rarely test that.

They do not test whether the person has seen production systems fail. They do not test whether they understand regulated data. They do not test whether they can sit with an owner and separate embarrassment from actual risk. They do not test whether they can simplify a plan until the client can run it without them.

That is judgment. Judgment is harder to certify.

The buyer's mistake

The mistake is not asking whether someone has credentials. The mistake is treating the credential as the answer.

A better question is: what did the credential teach you that changed how you work?

Another better question: show me a time you rejected an AI output that looked good.

Another: what data would you never put into a public AI tool?

Another: how do you test an AI-assisted workflow before it touches live client work?

Another: what would you talk me out of?

Those questions reveal more than a badge count.

A person who cannot answer them plainly may still have certificates. A person who can answer them with specifics is showing you something closer to operating experience.

Why I still earned them

So why bother?

Because I do not want to dismiss what I have not inspected.

AI changes quickly. Vendor documentation changes. Product names change. Privacy settings change. Certifications and learning paths are one way to force a structured pass through the material. They give me a reason to check what the vendor says now, not what I remember from three months ago.

They also help me explain the difference between tool knowledge and business judgment.

When I say certificates are limited, I am not saying that from outside the room. I am saying it as someone who holds them and still would not let them be the lead argument for why you should trust me.

The lead argument is the work: twenty years in environments where technology failure had consequences, daily use of AI tools, real review of outputs before they touch business systems, and a willingness to slow a client down when the tool is getting ahead of the decision.

The certificates sit behind that. They do not replace it.

How to read credentials

Treat AI credentials as signals, not proof.

A completion certificate says the person finished material. Useful, but light.

A proctored exam says the person met a defined standard under test conditions. Better, but still narrow.

A hands-on assessment, if one exists, says more. It still needs context.

Current daily practice matters. A certificate from last year in AI may already need updating.

Operating history matters more. Has the person made decisions where failure had consequences? Have they reviewed real systems? Have they protected sensitive data? Have they documented work someone else had to run?

And humility matters most. Anyone selling certainty in this field is telling you something important about their own risk tolerance.

The right advisor should be able to say, "I know this," "I need to verify this," and "we should not do that" with equal comfort.

That is worth more than a stack of certificates.

If you want help judging AI by more than badges, start a conversation.