People
Human Certification
Your team trains on the standard, then demonstrates it when the conversation gets difficult.
For your people →Certification
Prove they’re ready.
Confidence that critical work keeps moving: evidence your people and AI agents can apply the judgment the work requires, even when the people it used to depend on are not in the room.
Who gets qualified
People, AI agents, trainers. One expert standard.
AI agents
AI Agent Certification
Your agent is evaluated on the decisions that matter in your environment, not on fluent answers.
For your agents →Trainers
Trainer Certification
Qualify the people who teach the standard. Scoped with your program.
For your trainers →What you receive
Evidence you can act on.
01
Assessment
Scenario performance against the expert standard.
02
Failure analysis
Where judgment broke, and whether it was critical.
03
Environment scope
Exactly where the qualification holds.
04
Certification record
Certified, or refused with reasons. Versioned.
05
Revalidation plan
When and why it must be re-earned.
The first certified agent
Katya is what a certified agent looks like.
Katya, the first certified Kataclyzim agent, operates from expert-derived Resistance Intelligence. Her version, environment and scope are readable in her public record.
Four tracks, one standard
Most of the risk lives in the handoff. So that gets qualified too.
A person can be ready and an agent can be ready and the work can still fail between them — the agent escalates late, the human inherits no context, or both assume the other is handling it. Human + AI qualification tests the shared work, not just the two parties separately.
Track 01
Human
A cohort trained and tested on the judgment the role requires. The manager gets readiness, not attendance.
Human Certification →Track 02
AI agent
Bring your own agent, or Katya. Tested on scope, boundaries and escalation before it is allowed near a customer.
AI Agent Certification →Track 03
Human + AI
The shared workflow: when the agent must hand over, what the person receives with it, and who owns the decision at each point.
Scope a dual program →Track 04
Trainer
Qualify your own people to run the program internally, so the capability stops depending on us as well.
Trainer Certification →What the dual track actually tests
Whether the agent recognises the point at which it should stop and escalate rather than continue. Whether the person receiving the handoff can read the situation the agent could not. Whether the record of what happened is good enough for the next person to pick it up. And what each side is cleared to decide alone.
What you are buying
Nine things arrive in every qualification program.
Scope and depth are set in the engagement. The shape does not change.
01
The governing standard
Which Judgment Set, which version, and the environments it covers. Named on the record.
02
Preparation
Training and practice against the standard before assessment, so the test measures readiness rather than surprise.
03
Realistic scenarios
Situations from the subject's own environment, including the uncomfortable ones and the ones where stopping is correct.
04
Assessment
A scored evaluation against the standard. The evaluation method itself is not disclosed; what it concluded is.
05
Failure analysis
Where the read went wrong and why — the pattern behind the miss, not a list of incorrect answers.
06
A scoped decision
Cleared for which work, in which environments, with which limits. Never a general-purpose pass.
07
The evidence record
What the organisation keeps: subject, standard version, scope, outcome, conditions and date. Auditable internally.
08
Deployment conditions
What has to be true in the live operation for the clearance to hold — supervision, escalation paths, volume limits.
09
A revalidation plan
The triggers and the cadence that keep the proof current instead of historical.
The operational decision
A qualification is a decision, not a certificate.
There are five outcomes, and four of them are not a pass. Each one tells the business something it can act on this week.
Certify
Cleared for the named work in the named environments, under stated conditions.
Restrict
Cleared for part of the work only — a narrower environment, supervised handling, or lower-stakes cases first.
Refuse
Not cleared. A refusal is a finding with reasons, and for an agent it is the cheapest one you will ever get.
Remediate
Targeted work on the specific gap the analysis found, then a re-test on that scope.
Revalidate
The work changed, so the proof is re-established against the current standard version.
Practice screens and roleplay are preparation. They are not evidence, and nothing in a practice session becomes a qualification record.