Frame
Know what problem is actually being solved.
Define the decision, the stakeholders, the objective, the constraints, the evidence required and what success looks like.
Bad agentic work often starts with a poorly framed task.
The Agentic Actuary
Manifesto.
More valuable work. Deeper understanding.
Judgement that stays your own.
Our mission
To develop actuaries who can use agents to do substantially more valuable work without giving up understanding, judgement or professional responsibility.
The goal goes beyond learning to use ChatGPT. We are developing a new professional capability.
An Agentic Actuary can direct powerful AI systems, interrogate their work, make the final judgement and defend the outcome. That is the standard every level is built around.
The graduate promise
Someone who earns the accreditation should be able to say:
“I can take a real actuarial problem, use agents to investigate it deeply, verify what matters, make my own professional decision, and explain why that decision deserves reliance.”
That is the credential. Everything else is curriculum.
This manifesto sets the standard we are building towards. Formal accreditation remains in development.
The six capabilities
The whole accreditation is built around six capabilities, developed together and demonstrated through real actuarial work.
Direct
Use agents as capable collaborators, not answer machines.
Decompose work, delegate meaningful tasks, choose appropriate tools, coordinate analyses and iterate intelligently. Keep the project moving towards the objective.
This is much broader than prompt engineering.
Investigate
Use evidence, not plausibility.
Obtain and interrogate the right data. Test explanations, compare alternatives, run calculations, explore sensitivities and distinguish findings from assumptions.
Agents expand what can be investigated.
Verify
Know what must be checked before relying on the work.
What supports this? What does it depend on? What could make it wrong? What should be tested next? Verify important calculations, challenge assumptions and look for missing explanations.
This is where Acta becomes fundamental.
Explore ActaJudge
Make the decision yourself.
Synthesise evidence and uncertainty into a proportionate professional recommendation.
The agent can advise. The actuary owns the judgement.
Defend
Take responsibility for the conclusion.
Explain, without AI, why you reached the decision: the evidence and assumptions that mattered, the uncertainty, the alternatives you rejected and what would make you change your mind.
If you cannot defend it, you do not yet own the work.
The core way of working
- Frame
- Direct
- Investigate
- Verify
- Judge
- Defend
In teaching, this becomes Investigate → Decide → Defend. The six capabilities remain the underlying framework.
Our principles
AI should increase capability without decreasing understanding.
Producing more work while understanding less of it does not make someone a better actuary.
The agent does not own the judgement.
The professional remains accountable for the decision.
Plausible is not the same as reliable.
Good agentic work is grounded in evidence, tests and appropriate challenge.
Verification is part of doing the work.
It is not a final review step added after the agent finishes.
The path matters, not only the answer.
Two candidates can reach the same recommendation while demonstrating very different levels of professional capability.
Uncertainty should remain visible.
Good professionals know what they do not know.
Agents should expand the questions an actuary can investigate.
The goal goes beyond doing the old work faster: deeper, broader and more decision-useful work.
The human must be able to defend the outcome without the machine.
That is the ultimate test of ownership.
What the accreditation is not
- A prompt-engineering qualification.
- An AI literacy course.
- A test of which model someone uses.
- A competition to generate the best report.
- A certification that someone can automate everything.
- A replacement for actuarial knowledge.
It is a standard for the ability to practise actuarial science effectively in an agentic world.
Operate. Integrate. Lead.
Three levels of increasingly demanding professional work.
Level I
Operate
Can you work effectively with agents on a bounded actuarial problem?
Frame it, investigate it, verify key work, decide and defend.
Level II
Integrate
Can you handle ambiguous, multi-stage professional work with agents?
Competing explanations, changing evidence, multiple analyses, uncertainty and trade-offs.
Level III
Lead
Can you exercise senior professional judgement in an agentic environment?
Multiple agents, conflicting objectives, governance, high-stakes decisions, strategic recommendations and accountability.
The standard behind everything.
The Agentic Actuary Accreditation is being developed to certify that an actuary can use AI to do more powerful work while retaining the judgement, understanding and accountability that make the work professionally reliable.
Begin with Level I