Certifications For AI Consultants: What Is Still Current

Start with an audit rather than a list, because two of the credentials most commonly recommended to AI consultants cannot be taken any more.

AWS Certified Machine Learning – Specialty is retired. AWS states on its own certification page that the last day to sit the exam was 31 March 2026, and that existing holders keep an active certification for three years from the date they passed.

The TensorFlow Developer Certificate exam is closed. TensorFlow’s certificate page states that the programme has been closed while the team evaluates its next step, and that credentials already earned remain valid for three years from the pass date.

Both still appear across current recommendation articles, including on this page before this update. That is the first thing worth knowing about certification advice: it ages faster than the credentials themselves, and nobody goes back to check.

Retired credentials still dominate the recommendation lists

The third row is the one that keeps recurring. Salesforce renamed its Data Cloud Consultant credential to Data 360 Consultant and its Administrator credential to Platform Administrator, and both old URLs now redirect. A certification list written before a rename does not become wrong so much as unrecognisable, which is worse for someone searching for study material.

Certifications sort into four jobs and most lists cover only one

The deeper problem with the standard list is not staleness. It is that six recommendations turn out to be six versions of the same credential: build a model on somebody’s cloud. Consultants need coverage across categories, not depth in one.

Foundational literacy proves you can hold an informed conversation about AI without building it. Cheap, fast, and the right first credential for anyone whose job is advising rather than implementing.

Platform engineering proves you can build and deploy on a specific cloud. Valuable, expensive in study time, and only relevant where clients actually run that cloud.

Ecosystem and agent work proves you can configure and govern agents inside an application platform your client already owns. This is the category that grew fastest as the work moved from building models to deploying vendor agents, and it is missing from almost every generic list.

Governance and law proves you understand accountability, risk and the regulation applying to AI systems. Almost nobody recommends it, and it is increasingly the thing buyers ask about.

Most consultants need one from foundational or platform, plus one from ecosystem or governance. Four from a single category signals depth in a narrowing skill rather than range.

AWS Certified AI Practitioner is priced for people who do not build

The foundational tier is worth examining closely because it is the least understood. AWS positions this credential explicitly at people who use rather than build AI systems, and it names the roles: business analyst, IT support, marketing, product or project manager, line-of-business or IT manager, and sales.

Two details reward attention. The price sits at roughly a third of a specialty exam, which changes the calculation for consultants who want credible AI literacy without a modelling career. And the automatic recertification path means a foundational credential can be maintained by progressing rather than by repeating, which is unusual and quietly generous.

Governance is the category consultants skip and buyers ask about

When AI systems only made predictions that a human then acted on, accountability sat with the human. Agents that take actions moved that line, and the questions arriving in procurement now are about who is responsible when an agent does something wrong, what documentation exists, and which regulation applies.

The IAPP Artificial Intelligence Governance Professional credential covers exactly that territory: AI systems and their impacts, how current and emerging laws apply, the AI life cycle and the context in which its risks are managed, and the implementation of responsible governance.

For a consultant, its value is positional rather than technical. It is the credential that lets you be in the room for the conversation that decides whether a project proceeds, which is a different room from the one where the build is scoped. Very few AI consultants hold anything in this category, which is precisely what makes it worth holding.

Ecosystem credentials outrank generic ones inside the client’s own platform

A generic cloud AI certification tells a client you can build AI somewhere. An ecosystem credential tells them you can build it in the system they already run, against their data, their permissions and their release cycle. When the client has already chosen a platform, the second claim is worth considerably more.

In the Salesforce ecosystem the relevant credentials are the Agentforce Specialist, aimed at managing and optimising Agentforce with a working knowledge of platform configuration, and the Data 360 Consultant, aimed at the data pipeline underneath it. Neither appears on generic AI certification lists, and both are more directly useful to a consultant delivering agent work than a modelling credential is.

The general rule holds outside Salesforce too. Certify where the client’s licence spend already sits, because that is where your recommendation has to survive.

Certification renewal is the cost the lists never mention

Every credential above expires, and the expiry is the part that turns a certification portfolio into a standing obligation. AWS certifications run three years. The TensorFlow credential ran three years and can no longer be renewed at all, which is the most instructive case in this article: holders are now carrying a credential with a countdown and no path forward.

The practical consequence is that a large portfolio is a recurring commitment rather than a permanent asset. Nine certifications means nine renewal cycles, each with study time attached, and most consultants quietly let the majority lapse. It is more defensible to hold three current credentials across different categories than nine of which five have expired, since a lapsed certification on a profile invites exactly the question you do not want asked.

Before booking anything, work out what the renewal looks like. Some vendors offer free maintenance modules, some require the full exam again, and at least one now offers nothing at all.

Three questions decide which AI certification to take

The older version of this page listed six factors to weigh. In practice the decision collapses into a short sequence.

1. Do you build or advise? Advisors take foundational plus governance. Builders take platform plus ecosystem. Getting this wrong is how people end up studying MLOps for a job that never touches a pipeline.
2. Whose platform do your clients run? Certify there. A credential in a cloud none of your clients use is study time converted into nothing.
3. What will this look like in three years? Check the renewal terms and check whether the underlying product still carries the same name. Both questions take five minutes and both have caught out people who skipped them.

Where those answers sit within a longer trajectory, and which credentials matter at which stage, is covered in the AI consultant career path.

AI certification strategy goes wrong in six ways

The last row is the cheapest gap to close and the one with the clearest return, because governance credentials are still rare enough to be distinguishing rather than expected.

AI consultant certifications, condensed

Certification proves knowledge, delivery proves the rest

A credential clears screening. What a client remembers is whether the work landed, which is why the strongest consultant profiles pair a small number of current certifications with a visible record of delivery.

That record accumulates faster when the routine artifacts of an engagement stop consuming the week. GetGenerative.ai agents produce discovery, design, build, testing and support artifacts inside the platform, and consultants can try every agent free for a week to see what that leaves room for.

Questions consultants ask about AI certifications

Which AI certification is best for consultants in 2026?

There is no single best one, because the categories serve different jobs. Advisors are usually best served by a foundational credential such as AWS Certified AI Practitioner paired with a governance credential. Builders are better served by a platform engineering credential paired with an ecosystem one matching their clients’ stack.

Is the AWS Machine Learning Specialty certification still available?

No. AWS retired it, with the final exam date of 31 March 2026. Existing holders keep an active certification for three years from the date they passed. Anyone looking for an AWS machine learning credential now should look at the Machine Learning Engineer Associate.

Can I still take the TensorFlow Developer Certificate?

No. TensorFlow has closed the certificate exam while evaluating what comes next. Credentials already earned remain valid for three years from the pass date, with no renewal route currently published.

Do AI certifications expire?

Most do, commonly on a three-year cycle. Renewal terms vary considerably, from free maintenance modules to sitting the full exam again, and in at least one case there is now no renewal path at all. Check the terms before booking rather than after.

How many certifications should an AI consultant hold?

Two or three current ones across different categories is more persuasive than a long list where most have lapsed. Volume signals collecting; range signals that you understand where your work actually sits.

Are vendor certifications better than independent ones?

They answer different questions. Vendor credentials prove you can operate inside a specific platform, which matters most when the client already owns it. Independent credentials, particularly in governance, prove something that travels across platforms and tends to age more slowly.

Do certifications increase consulting rates?

Indirectly. They help you reach conversations rather than win them, and rate movement follows the work you can point to. The stage-by-stage view of what actually drives progression is a separate discussion from which exam to book.