Launching soon. CAI-1 “AI Foundations” will be free, for everyone. Launching soon · CAI-1 is free See the credential ladder →
Certavant

The Standard

The standard, in the open.

Four levels, one honest test, and everything we publish about how the grading works. This is the page we’d want to read before trusting any certification.

The CAI ladder

Four levels, from fluent to architect.

CAI-1

AI Foundations

Free
  • Proves you canWork with AI fluently — prompt with intent, verify what it gives you, and handle data responsibly.
  • AssessmentScored, hands-on exercises built into the lessons. Genuinely failable — a real screen, not a participation badge.
  • Who it’s forAnyone who uses AI and wants a real baseline to build on.
CAI-2

AI Practitioner — Role Track

  • Proves you canDo the AI-augmented work of a specific role, end to end.
  • AssessmentReal work products, produced in an AI workspace and graded against a published rubric, with your AI-collaboration process captured.
  • Tracks at launchMarketing · Sales · Customer Support · HR · Data & Business Analysis.
CAI-3

AI Automation Specialist — Role Track

  • Proves you canDesign, build, and oversee multi-step AI and agent workflows.
  • AssessmentAn automation project and capstone artifact, built with human-approval gates.
  • Who it’s forPeople whose job is becoming “supervise the agents.”
CAI-4

AI Systems Architect

Proctored capstone
  • Proves you canArchitect and govern AI systems across an organisation.
  • AssessmentA proctored capstone plus a short live oral-defence checkpoint — the highest-rigour credential we issue.
  • Who it’s forSenior practitioners and leaders accountable for how an organisation uses AI.

CAI-1 and the CAI-2 tracks lead the way in; the higher rungs roll out as the standard grows.

The assessment

How the exam actually works.

01

You get a real task.

Not questions about the work — the work itself. A campaign brief to produce, a messy dataset to fix, a set of support tickets to resolve. Each candidate gets a parameterised variant, so no two exams are identical.

02

AI is allowed and expected.

You work in a Certavant AI Workspace — a sanctioned AI chat panel plus an editor and file attachments. The workspace captures your entire process: every prompt, iteration, and verification step.

03

Your work is graded against a published rubric.

An AI grader scores your work product and your process against the same blueprint we published before the exam. No hidden criteria.

04

A named human examiner signs the decision.

The AI grader recommends; a qualified, named human examiner reviews the work and evidence and makes the final call. Every decision is logged, and every adverse one can be appealed to a fresh human re-grade.

05

Your credential is issued with the work embedded.

You get a public verify page at cert.education/v/… carrying your credential and the actual graded work as evidence.

Open-AI by design

Yes, you can use AI during the exam. That’s the point — and it’s harder, not easier.

An exam that bans AI tests whether you can do the job the slow way, alone. Nobody does the job that way anymore. So we did the opposite: we let you use AI, and we grade the things that actually separate good AI work from bad.

DelegationDid you ask the model for the right thing, with the right context?
JudgmentDid you catch where it was confidently wrong, and fix it?
VerificationCan you tell a solid answer from a plausible-sounding one?

Memorisation doesn’t help you here. Prompting a model to write your answer doesn’t either, because a fluent wrong answer fails a rubric that checks whether the work is actually right. Letting you use AI isn’t a loophole. It’s what makes the exam an honest test of the job.

Transparency commitments

What we publish about our own exams.

Most certifications ask you to trust the badge. We’d rather show our work.

Blueprints

Competencies, task types, scoring criteria, and the pass standard before you sit.

Pass rates

Per credential, updated over time.

Reliability

How closely AI grading agrees with expert human graders.

Human decision

A named examiner on every logged and auditable decision.

Appeals

A fresh human re-grade for adverse decisions, with overturn rates published.

Published methodology, independent decisions, and a paper trail you’re allowed to inspect.

A

Credentials that stay current

An AI credential from two years ago is a fossil.

The tools, models, and good practices all turn over fast. Every CAI credential carries an expiry, and renewal is a short, assessment-only refresh — you re-prove the current skill, not re-take the whole course. A live credential means current capability.

B

Challenge exam

Skip the course. Prove it directly.

If you already do this work well, take the same graded, open-AI assessment without the coursework in front of it. Pass, and you hold the same credential. An existing AI badge from another provider qualifies you to go straight to the challenge exam.

For employers

Verifying a credential takes one link.

You’re not trusting a logo. You’re reading the evidence.

cert.education/v/8ktq-2m9r

Verified

CAI-2

AI Practitioner — Marketing

Issued
March 2026
Expires
September 2027
Decision
Human examiner

Graded evidence

Campaign brief — product launch Open
Verification log — seeded errors Open
AI collaboration transcript Open

Signed by a named human examiner · Appealable

Illustrative example of a Certavant verification page

Every Certavant credential is a public page at cert.education/v/… showing the level, track, issue and expiry dates, current validity, and — where the holder makes it visible — the graded work behind it.