The three AI certifications worth having if you don't write code

Most AI certification advice is written for engineers, which is why it keeps recommending exams with a coding prerequisite to people who manage delivery rather than write it. There are now three foundational certifications aimed squarely at business audiences, none with prerequisites, all open to anyone who wants to pay. Here's the honest version of each.

The short answer

If you want one: Google Cloud's Generative AI Leader. It's $99, ninety minutes, valid three years, and roughly half of it is genuine AI literacy rather than product tour.

If your organization runs on Microsoft 365: AI Transformation Leader (AB-731) is the most precisely aimed at your job, and it's a forty-five-minute exam.

If you want the name with the broadest recognition: AWS Certified AI Practitioner, accepting that it's the most technical of the three.

Microsoft: AI Transformation Leader (AB-731)

Microsoft files this one under Role: Business Leader, Level: Beginner, and the audience description is unusually direct. Business decision-makers at any level guiding AI adoption, not expected to write any code.

Three domains:

Domain	Weight
Business value of generative AI	35–40%
Microsoft's AI apps and services	35–40%
Implementation and adoption strategy	20–25%

Note the middle third. That's Copilot, Copilot Studio, Microsoft Graph, and Foundry Tools. Genuinely useful if your organization is a Microsoft shop, and largely wasted effort if it isn't. Be honest with yourself about which one you are.

The first domain is the part that transfers anywhere: the difference between generative AI and other kinds, cost drivers including tokens and ROI, fabrications and reliability and bias, prompt engineering, retrieval-augmented generation, and the effect of data quality on outcomes. None of that is Microsoft-specific.

The third domain is the one I'd argue is under-weighted at 20–25%, because it's the part most projects actually fail on. Establishing governance, standing up an adoption team, naming the common barriers, running a champions program, and understanding the license models well enough to forecast a bill.

Practical details. Forty-five minutes, English only at the moment. Pass mark 700. Scheduled through Pearson VUE, and priced by the region you sit it in rather than a flat published fee. On expiry, be careful what you read: the study guide links to Microsoft's annual renewal process, which is a free open-book assessment on Microsoft Learn rather than a paid re-sit, but that same link appears on certifications that don't expire at all, and Microsoft's renewal page says the annual cycle covers associate, expert, and specialty credentials. AB-731 is none of those. Check your own certification profile once you've passed rather than trusting either of us on it.

One genuinely useful thing Microsoft does that the others don't: the skills-measured list carries a dated change log showing exactly what moved and when. If you're preparing over several weeks, check it before you sit.

Google Cloud: Generative AI Leader

The most accessible of the three, and the least locked to its vendor. Google's own description is "anyone in any job role, with or without hands-on technical experience," which is about as broad as these things get.

Four sections. Google publishes the weights as approximate rather than exact:

Section	Weight
Fundamentals of gen AI	~30%
Google Cloud's gen AI offerings	~35%
Techniques to improve model output	~20%
Business strategy for a gen AI solution	~15%

Add sections one and three together and you have half the exam on vendor-neutral material: what these systems are, and how to get better results out of them. That's the highest proportion of transferable content of the three, and it's why I'd start here even if you never touch Google Cloud.

Practical details. $99 plus tax. Ninety minutes, 50–60 multiple-choice questions. Online-proctored or at a test center. No prerequisites. Valid three years, the longest of the three, which is worth weighing if you dislike renewal admin.

AWS: Certified AI Practitioner (AIF-C01)

The most widely recognized name, and the most demanding of the three.

Five domains:

Domain	Weight
Fundamentals of AI and ML	20%
Fundamentals of generative AI	24%
Applications of foundation models	28%
Guidelines for responsible AI	14%
Security, compliance, and governance	14%

AWS describes the target candidate as someone who uses rather than necessarily builds AI solutions, and names business analysts, marketers, product and project managers, line-of-business managers, and sales professionals among the roles it's for. That's accurate as far as it goes. But look at the weights: applications of foundation models at 28% is the single heaviest domain across all three exams, and it goes further into how these systems are assembled than either of the others does.

If you have no technical background at all, take this one third rather than first.

Practical details. $100. Ninety minutes, 65 questions, of which 50 count toward your score and 15 are unscored trial items you won't be able to identify. Pass mark 700 on a 100–1000 scale. No prerequisites. Valid three years.

The order I'd take them in

Google first. Cheapest, broadest, and the highest share of knowledge that transfers. If you only ever do one, this is the one that makes you more useful in the widest range of rooms.

Microsoft second, if and only if your organization actually runs on Microsoft 365. The middle domain is a third of the exam and it's a product tour: valuable when you use those products daily, dead weight when you don't. The forty-five-minute format is in its favor either way.

AWS third. By that point the first two domains will feel like revision, and you'll have the grounding to handle the heavier applications material without it turning into a slog.

What none of them will teach you

All three are multiple-choice exams about concepts. None of them will teach you how to tell whether the AI feature your team just demoed is actually working, how to argue a budget for something whose cost scales with usage, or what to do when the pilot succeeds and the rollout stalls.

That's not a criticism of the certifications so much as the limit of the format. Use the exam as the deadline that makes you learn the material. Then do the harder work of applying it, which is where the credibility actually comes from.

A note on a fourth option

You may also come across Anthropic's Claude certifications, including an Associate-level exam explicitly aimed at non-technical business professionals. It's a good blueprint. The catch is registration: it requires membership in the Claude Partner Network and a company-domain email address. Joining the network is free and open to any organization bringing Claude to market, but a personal email won't work, and if your employer isn't a partner you can't simply pay and sit it.

Worth knowing about. Not worth planning around unless your organization is already in that network.