Not every certification is worth the hours you'd put in. If you haven't already, read Which Jobs Are Most Exposed to AI first. Reskilling makes most sense once the trigger points from your income backup plan have started going off.

Postings
the only test that matters: is it named in live job ads?
20+
live postings before a credential is a hiring signal
Project
applied work beats a multiple-choice certificate
Payback
total cost ÷ the pay difference the postings advertise

There are no percentages here, on purpose. Figures on reskilling returns, employer plans and pay rises get passed around a lot but rarely trace back to data you can check, and they vary hugely by field and country. The tests above are things you can check yourself in an hour, which is more use than a statistic you can't.

Don't wait to be trained

You'll often see a pair of statistics on this: that most business leaders now call AI skills critical, while only a tiny minority of companies have started serious retraining for their existing staff. The exact figures vary from survey to survey and are hard to check, so be wary of the specific numbers.

A paper collage of a stack of books with a rolled certificate, a three-step staircase and a caption reading NEW SKILL

The general picture does match what most people see at work, though: employers talk about AI skills a lot more than they pay for training in them. So waiting to be trained is a weak plan, simply because your employer has less reason to train you than all the talk suggests. If there is a training budget, ask for a specific, named course tied to a specific business result rather than "training" in general.

What separates a certification that gets you hired

Certifications that make you build something real tend to be worth more than ones that just test you with multiple-choice questions. Ask three questions before you start any course:

  1. Does it end with something concrete, like a project or a portfolio piece, or just a test?
  2. Is it from a provider employers in your field actually recognise?
  3. Can you find real job adverts in your field that name this certification?

Those questions are easy to nod along to and just as easy to skip, so here's how to actually run each one, in about an hour all told.

Run the job-advert test properly

This is the check that saves the most wasted months, and almost nobody does it before enrolling rather than after.

  1. Open whichever job board is biggest in your field and search for the exact certification name in quotes, limited to your country or region.
  2. Count the results. If there are fewer than about twenty live adverts, the qualification isn't yet something employers hire on where you live, whatever its reputation online.
  3. Read ten of them properly. You're checking whether the certification appears under requirements or just under "nice to have", and that difference is the whole point.
  4. Check the salaries are ones you'd actually move for, and that the adverts are recent rather than reposts that have been open for months.
  5. Save three adverts. They become your brief: work towards what those roles ask for, not just towards finishing the course.

If a certification passes, you've also found your target roles. If it fails, you've saved yourself a few hundred pounds and several months for the price of an hour's searching.

What "applied project" really means

Course marketing uses "hands-on" and "project-based" very loosely. A guided exercise where you follow along and everyone ends up with identical results isn't a portfolio piece. A hiring manager has seen that same project forty times and it tells them nothing about you.

A real applied project has three things going for it: you choose the problem or the data, you have to make design decisions that could reasonably have gone another way, and you end up with something you can show and talk through in an interview. If a course doesn't offer that, you can usually add it yourself by redoing the final project on a problem from your own industry, which also plays to the knowledge you already have.

Budget the time and money against your runway

Reskilling is an investment that has to pay itself back, so write the numbers down before you commit rather than finding them out halfway through.

  • Count real hours, not advertised ones. Take the course's stated hours and add about half again for the project work and the bits you'll need to redo. Then check that honestly against the evenings you actually have free.
  • Count the total cost, not the sticker price. Include exam fees, any renewal costs and any software or cloud costs the project needs.
  • Work out the payback. Against the pay rise those saved adverts actually offer, how many months until it's paid for itself? If that's longer than you'd realistically stay in the role, it's the wrong course.
  • Think about your runway. If you're paying for this out of your cash cushion rather than spare income, it eats into the months of cover you'd have if things went wrong. That's sometimes the right call, but make it on purpose.

Order it so you're employable partway through

Long programmes can leave you with nothing useful until everything's finished. Avoid that by ordering the work so each stage leaves you more hireable than the last. Learn whatever appears in the most adverts first, build your project as early as you can rather than saving it for the end, and put the project somewhere public early on. Six weeks in with one project you can show beats six months in with a half-finished syllabus, and it means a surprise redundancy doesn't leave you stuck mid-course with nothing to show for it.

What to skip

  • Certifications with no applied project
  • Anything not named in real job adverts in the field you're aiming for
  • Generic "AI fundamentals" courses aimed at a mass audience
  • Programmes whose main selling point is the platform's brand rather than what you learn
  • Anything that needs an expensive annual renewal to stay valid, unless your field genuinely requires it
  • Courses that promise to place you in a job without naming the employers

Reskilling decision checklist

0 of 5 done

Frequently asked questions

Are AI certifications actually worth it in 2026?

Some are, and the difference is predictable. Certifications that end in an applied project you chose, from a provider employers in your field recognise, and that appear by name in live job postings where you live, tend to pay back. Ones that end in a multiple-choice test and aren't named in any posting generally don't, regardless of how heavily they're marketed. The one-hour job-posting check above separates the two before you spend anything.

How do I know if employers recognise a certification?

Search the exact certification name in quotes on the dominant job board in your field, filtered to your country. Fewer than about twenty live postings means it isn't yet a hiring signal in your market. Then read ten postings and check whether it sits under requirements or merely under "nice to have", that distinction matters far more than the total count.

Will my employer pay for or provide reskilling?

Ask, but don't plan around it. Surveys consistently show employers rating AI skills as critical far more often than they fund training for existing staff, though the precise figures differ by survey and are hard to verify. If a training budget does exist, the practical move is to ask for a specific named course with a named business outcome rather than for "training" in the abstract.

How long should reskilling take before it pays off?

Work it out rather than guessing: total cost including exam fees, renewals and any software, divided by the monthly pay difference the postings you saved actually advertise. If the payback period is longer than you'd plausibly stay in the role, it's the wrong course. Budget the time the same way, take the advertised hours and add roughly half again for the applied work and the parts you'll redo.

Do I need to learn to code to stay employable?

For most roles, no. What's rising in value is the ability to use AI tools well within a domain you already understand, judging output, spotting where it's wrong, and knowing what "good" looks like to the person signing off. Deep domain knowledge plus competent tool use is a stronger and much faster position than starting from zero as a generic developer, and it builds on what you already have.

When is the right time to start reskilling?

While you're still employed and not under pressure, because that's when you can afford to be selective and fund it from surplus income rather than from your emergency cushion. If you've written trigger points as part of an income backup plan, two firing in the same quarter is a reasonable prompt to commit to something specific rather than continuing to research options.

Where to go next

This page is informational, not career advice. Certification outcomes vary by field, region, and effort.