AI Guides › Playbooks

The Evidence-Before-Certificate Ladder

By Nigel Guy · 6 min read

Most "get AI certified" advice ends with you holding a PDF and a badge for LinkedIn. It feels like progress because a certificate is easy to collect and easy to show. The trouble is that a hiring manager, or a client, is asking a different question: can this person do the work, and can I see it? A badge answers that only weakly, and only if the issuer and what was tested are known to the reader.

The rule: spend your first fortnight producing evidence of work you can show, and buy or earn a certificate only where a named reader has asked for it.

The mechanism: the Evidence-Before-Certificate Ladder

Four rungs. Climb in order, and stop as soon as the next rung stops being useful to your actual target reader.

Rung What you hold What it tells a reader Typical cost
1. Working proof One to three real tasks done with AI, with before/after and what you checked You can do the job and you notice errors Free (your time)
2. Free structured learning A course you finished, ideally tied to a piece of rung 1 work You learned the vocabulary and limits Often free
3. Proctored or vendor exam A credential with a published exam, pass mark and issuer A third party tested a defined syllabus Exam fee, varies
4. Employer- or role-specific credential Something a job advert or client names You meet a stated requirement Varies

Most people start at rung 3 and never build rung 1. Reverse that.

What do you sort out before you start?

Write one sentence: "I want [ROLE OR CLIENT TYPE] to see that I can [SPECIFIC TASK]." Then find three real job adverts or client briefs for that target and copy out any line that names a certificate, tool or skill. If none of the three names a certificate, rung 3 is optional for you. If one does, you now have a reason, and a named credential to check.

What do you need instead of a shelf of certificates?

A proof page: one page (a document, a public portfolio page or a PDF) with up to three entries. Each entry has four lines: the task, what you gave the AI, what you changed or rejected in its output, and the result. The "what you changed or rejected" line matters most. It is the evidence that you exercise judgement, which is the part a certificate cannot show.

What's the method for the next fortnight?

  1. Days 1-2: pick the target. Do the one-sentence exercise above and collect the three adverts.
  2. Days 3-7: do two real tasks. Use work you already do, or a realistic sample with no confidential data. Keep your prompts and the outputs.
  3. Days 8-10: take one free course that matches a gap. Anthropic's free Academy, at time of writing, lists courses such as AI Fluency: Framework and foundations (about four hours) and AI capabilities and limitations (about three and a half hours). The page I checked did not mention certificates of completion, so treat it as learning, not a credential. OpenAI also runs an Academy with courses by audience; I could not confirm from its page what, if anything, it certifies or costs, so check that yourself before relying on it.
  4. Days 11-12: write the proof page. Three entries maximum.
  5. Days 13-14: decide on rung 3. Only if your adverts name one.

How do you check a course or credential before you pay?

Run this checklist. Every "no" is a reason to hold your money.

Question Where to look Pass looks like
Who issues it? The issuer's own site, not the seller's A named organisation you can verify
Is there a proctored exam or only a quiz? Exam page or FAQ A published exam with a stated pass mark
Does it name the syllabus? Study guide or skills outline Percentages or objectives you can read
Does it expire? Renewal page A clear yes or no
Does any advert you collected name it? Your three adverts At least one
What is the exact £ price, with tax? Checkout A figure you have seen, not a guess

What does this look like in practice?

A hypothetical example: Priya, an operations coordinator, wants to move into an AI-enabled operations role. Two of her three adverts say "familiarity with generative AI tools" and none names a certificate. She spends her fortnight on two entries: a rewritten supplier-onboarding checklist and a weekly report template, each with notes on what the AI got wrong. She takes the free capabilities-and-limitations course for the vocabulary, and decides against paying for an exam. Her proof page is the thing she sends with applications.

Now a different hypothetical: Sam applies to a cloud-focused employer whose advert asks for Microsoft Azure credentials. Here rung 3 is justified. Microsoft's current exam for this level is AI-901, Microsoft Azure AI Fundamentals, which the exam page says suits people at the start of AI solution development and expects some Python and Azure familiarity. It is split roughly 40-45% AI concepts and 55-60% implementing AI solutions with Microsoft Foundry, with a passing score of 700. The page shows a price that depends on the country where you sit the exam, so check the £ price at booking. Microsoft states that Fundamentals certifications do not expire, whereas associate, expert and specialty ones need annual renewal (free, via an online assessment). Note the English exam was updated on 15 April 2026, so older study notes may be stale.

What are the traps?

How do you know it's working, and what won't this do?

It is working if, after the fortnight, you can send one link that shows three pieces of work and your reasoning, and if you can say why you did or did not take a credential. Success is not "I got interviews"; that depends on the market, your background and luck, and no one can promise it.

What to skip

Guardrails

Sources

All 751 AI guides · JulieMango plans from £17/mo