ScenarioGeneral-purpose autonomy assurance

General-Purpose Autonomy Assurance Architect

Also called AGI Assurance Architect

A General-Purpose Autonomy Assurance Architect would design how an organisation gains justified confidence in systems given broad, consequential work. Whether the role exists depends on how AI develops, and nobody is hiring for it now.

What they do

If broad reliable delegation became feasible, this person would define what a system may be trusted to do and what evidence would support that trust. They would map delegated authority across tasks and domains, set evidence requirements for each level of authority, and design checks that do not share failure modes. They would work with evaluation, security, legal and operations teams. They would review incidents, challenge weak evidence and revise the assurance structure as systems and uses change. They would also write the conditions under which delegation should be narrowed or withdrawn.

What a strong hire delivers

In the first months, a strong hire would deliver a clear map of what authority is delegated and to what. They would set out evidence requirements tied to each level of authority, with named owners. They would identify where evidence is thin or not independent, and propose fixes. They would also agree triggers for reducing delegation, so leaders know in advance what would change their minds.

How you can tell they are good

Ask them to take one delegated task and lay out the assurance case from scratch. Strong people separate evidence from assertion, say where their own evidence is weakest, and explain how each check stays independent of the system it tests. Ask what would make them recommend less autonomy. Worry about anyone who treats benchmark scores as sufficient or cannot name a failure.

Why it is hard to fill

If this role emerged, people would be scarce because it spans several fields that rarely overlap. Most candidates would be deep in one area, such as evaluation, control or systems safety, and thin in the rest. Titles would vary and none would match cleanly. No one would have assured broadly capable systems in live use, so judgement would rest on adjacent experience.

Where to find them

Today they sit in AI control research, model evaluation, AI governance and assurance, and frontier scenario analysis. Others come from systems safety in aviation, nuclear, medical devices or critical infrastructure, and from formal methods and security assurance. Look in frontier labs, safety institutes, independent evaluation groups, regulated engineering firms and policy research teams.

Interview questions for a General-Purpose Autonomy Assurance Architect

  1. Describe assurance work you have done where the system could cause serious harm. What evidence did you require and why?

    A strong answer names a real system, the claims being made, and the evidence behind each. They explain trade-offs and what they rejected. Worry if they describe only process, or cannot say what evidence was weak.

  2. How would you decide how much authority a system should hold over a consequential task?

    Good answers tie authority to consequence, reversibility and evidence quality, and describe graded limits. They mention monitoring and withdrawal. Be wary of answers that rest on capability alone or give a single threshold.

4 more questions, with what a strong answer covers, are in the Role Kit.

Job description you can post

We are looking for an architect to design how we gain justified confidence in AI systems that are given broad, consequential work. You would define delegated authority, set the evidence needed to support it, and keep our assurance independent and current as systems and uses change.

What you will do

  • Map the authority delegated to AI systems across tasks and domains.
  • Set evidence requirements for each level of authority.
  • Design checks that are independent of the systems they assess.
  • Work with evaluation, security, legal and operations teams.
  • Review incidents and revise the assurance structure.
  • Define triggers for narrowing or withdrawing delegation.
  • Report evidence gaps clearly to senior leaders.

What you will bring

  • Demonstrated assurance or safety case work on high-consequence systems.
  • Hands-on experience evaluating AI model behaviour.
  • Clear understanding of failure modes, including adversarial ones.
  • Evidence of holding a firm position under commercial pressure.
  • Ability to write clear evidence standards others can follow.
  • Experience working across technical and governance teams.

Nice to have

  • Background in AI control or formal methods.
  • Experience in a regulated safety-critical industry.
  • Work on scenario analysis for frontier systems.

Free to use and adapt. If it saves you time, a link back to this page is appreciated.

Depends on a scenario

Broad reliable autonomous delegation

What AI would need to do
Systems complete diverse consequential tasks with sustained reliability beyond existing narrow deployments.
What organisations would need to decide
Organisations explicitly delegate bounded authority and establish enforceable accountability.
What would rule it out
Persistent reliability limits or lack of useful cost advantage prevents broad delegation.

We give no date. The role becomes real if both conditions hold.

Skills that matter

  • AI controlAdvancedLimit harm from potentially untrusted AI behaviour
  • AI governance and assuranceAdvancedAssign accountability and verify lifecycle controls
  • Model evaluationWorkingMeasure capability and failure modes credibly
  • Frontier scenario analysisWorkingExplore conditional outcomes without claiming predictions

Other roles in general-purpose autonomy assurance

ScenarioCross-Domain Capability Certification DesignerA Cross-Domain Capability Certification Designer would define what a general-purpose AI system can credibly be said to do across consequential…ScenarioAutonomous Delegation Accountability OfficerAn Autonomous Delegation Accountability Officer would keep track of who authorised what when AI systems carry out broad work. The role depends on…ScenarioGeneral-Purpose System Decommissioning PlannerA General-Purpose System Decommissioning Planner would prepare the controlled withdrawal of autonomous systems that run deeply inside an…ScenarioHead of Autonomous OperationsA Head of Autonomous Operations would run a business unit staffed mostly by agents with a small human team. It depends on how AI develops: it only…

Related roles in other families

ScenarioRecursive Improvement Change-Control ArchitectRecursive AI development change controlA Recursive Improvement Change-Control Architect would define enforceable limits on fast, automated changes to AI systems. The role depends on how…ScenarioAutomated AI Research Assurance LeadRecursive AI development change controlAn Automated AI Research Assurance Lead would challenge the evidence coming out of heavily automated AI development. The role depends on how AI…ScenarioCapability Acceleration Monitoring SpecialistRecursive AI development change controlA Capability Acceleration Monitoring Specialist would watch for evidence that AI systems are improving unusually fast through automated research. It…ScenarioSelf-Modifying System Verification EngineerRecursive AI development change controlA Self-Modifying System Verification Engineer would check that stated safety properties hold as AI systems change their own software. The role…

How we wrote this

Role definitions are our own editorial work, checked against occupational classifications, standards, research and employer pages. Titles vary by company, and this is a description of the work, not a job advert. Scenario roles describe responsibilities that depend on stated conditions; they are not forecasts.

← General-purpose autonomy assuranceAll new job families

Start a search

Hiring a General-Purpose Autonomy Assurance Architect?

Tell us the work. We find the people already doing it, whatever their job title says, and you meet them already interviewed for your role.