Roles the authority layer will need
Seven role clusters, each labelled with the Institute's own assessment of its maturity. Emerging: already advertised, filled or legally required somewhere, under varied titles. Forecast: named in a framework, regulation or research whose effects fall in the next one to seven years. Proposed by the Institute: a title the Institute proposes, flagged as such. None of these is a recognised profession, an official qualification or an established professional title, and the Institute claims no title as its property.
| Role | Problem it solves | Authority or risk it handles | Evidence it produces | Where it may arise | Maturity |
|---|---|---|---|---|---|
| Mandate Engineer | Turns an instruction or business need into a bounded, verifiable, revocable mandate for a synthetic actor | Scope, permissions, prohibitions, limits, escalation and revocation for one or several agents | Mandate Specification and Authority Envelope; Claims-Evidence Matrix; run and revocation records; Evidence Record | Any organisation delegating consequential work to agents; the first role the Foundation programme prepares | Proposed by the Institute |
| Delegation Architect | Designs how authority propagates through orchestrators, sub-agents, tools and identities without losing attribution | Sub-delegation, machine identity, permission inheritance, separation of duties, nested oversight | Delegation and identity maps; permission and enforcement matrices; fleet mandate registers | Organisations running agent fleets or orchestrated multi-agent workflows | Forecast |
| Evidence Architect | Makes synthetic work verifiable by design: what must be logged, retained, hashed, redacted and reviewable | Provenance, negative evidence, immutability of criteria, retention, redaction | Evidence requirements, integrity records, audit trails a stranger can follow | Regulated sectors; audit; organisations whose agent outputs feed decisions | Forecast |
| Agent Evaluations Engineer | Establishes whether an agent behaves within mandate under normal, adversarial and failure conditions | Boundary tests, prohibited-action tests, missing-evidence tests, drift | Test plans and results; dry-run and adversarial records; evaluation reports | AI programme teams; assurance functions; vendors and buyers of agent systems | Emerging |
| Synthetic Enterprise Architect | Designs organisations in which human–agent work cells, fleets and machine-to-machine processes operate under coherent authority | Portfolio-level mandate design; organisational accountability; synthetic labour allocation | Enterprise mandate architecture; fleet topologies; governance decision records | Enterprises restructuring around agent groups | Proposed by the Institute |
| Agent Fleet Manager | Operates a group of agents day to day within mandates, budgets and escalation rules | Runtime limits, deviations, incidents, suspension and recovery across many actors | Run records, incident records, revocation cascades, operator notes | Operations functions adopting fleets; shared-service centres | Forecast |
| AI Liability Underwriter | Prices and bounds the risk of authorised synthetic action | Residual risk, control ownership, evidence quality, revocability, accountability | Underwriting assessments based on mandate and evidence quality | Insurers, reinsurers, risk and legal functions | Forecast |
The maturity labels are the Institute's own assessment, not an independently evidenced ranking. The Foundation programme prepares the first role; the others belong to the proposed Schools of Verification & Evidence, Synthetic Enterprise & Organisation and Agent Governance, none of which has faculty, learners or graduates.
Each role in one card
On narrow screens the matrix above already presents each role as a card.
Mandate Engineer
Solves: Turns an instruction or business need into a bounded, verifiable, revocable mandate for a synthetic actor.
Handles: Scope, permissions, prohibitions, limits, escalation and revocation for one or several agents.
Produces: Mandate Specification and Authority Envelope; Claims-Evidence Matrix; run and revocation records; Evidence Record.
Where it may arise: Any organisation delegating consequential work to agents; the first role the Foundation programme prepares.
Delegation Architect
Solves: Designs how authority propagates through orchestrators, sub-agents, tools and identities without losing attribution.
Handles: Sub-delegation, machine identity, permission inheritance, separation of duties, nested oversight.
Produces: Delegation and identity maps; permission and enforcement matrices; fleet mandate registers.
Where it may arise: Organisations running agent fleets or orchestrated multi-agent workflows.
Evidence Architect
Solves: Makes synthetic work verifiable by design: what must be logged, retained, hashed, redacted and reviewable.
Handles: Provenance, negative evidence, immutability of criteria, retention, redaction.
Produces: Evidence requirements, integrity records, audit trails a stranger can follow.
Where it may arise: Regulated sectors; audit; organisations whose agent outputs feed decisions.
Agent Evaluations Engineer
Solves: Establishes whether an agent behaves within mandate under normal, adversarial and failure conditions.
Handles: Boundary tests, prohibited-action tests, missing-evidence tests, drift.
Produces: Test plans and results; dry-run and adversarial records; evaluation reports.
Where it may arise: AI programme teams; assurance functions; vendors and buyers of agent systems.
Synthetic Enterprise Architect
Solves: Designs organisations in which human–agent work cells, fleets and machine-to-machine processes operate under coherent authority.
Handles: Portfolio-level mandate design; organisational accountability; synthetic labour allocation.
Produces: Enterprise mandate architecture; fleet topologies; governance decision records.
Where it may arise: Enterprises restructuring around agent groups.
Agent Fleet Manager
Solves: Operates a group of agents day to day within mandates, budgets and escalation rules.
Handles: Runtime limits, deviations, incidents, suspension and recovery across many actors.
Produces: Run records, incident records, revocation cascades, operator notes.
Where it may arise: Operations functions adopting fleets; shared-service centres.
AI Liability Underwriter
Solves: Prices and bounds the risk of authorised synthetic action.
Handles: Residual risk, control ownership, evidence quality, revocability, accountability.
Produces: Underwriting assessments based on mandate and evidence quality.
Where it may arise: Insurers, reinsurers, risk and legal functions.
Four roles at fleet scale
The proposed Agent Fleet Governance: Applied Case Lab and its ME-001 case address the work of four of the seven role clusters. Maturity labels are unchanged.
Delegation Architect
ME-001 Artefact C — the Delegation and Accountability Map: every edge with both accountable owners and its enforcement point; inheritance that narrows, never creates (scenario S1).
Agent Fleet Manager
Artefacts D and E — runtime authorisation decisions under aggregate limits and change-of-plan events; incident, revocation cascade and verified recovery (scenarios S2, S3, S5).
Evidence Architect
Artefact F — the Claims-to-Evidence Matrix with a falsifier for every claim; reconstruction of an incident from fleet evidence and a stated boundary of proof (scenarios S4, S6).
Synthetic Enterprise Architect
Artefact G — the Independent Governance Review with a continuation decision; the bridge to the proposed Level 3 programme on enterprise-level decision rights and risk concentration.
None of these is a recognised profession or an official qualification; the Lab is a proposed programme that has not run. Applied Case Lab
Programme enquiries
The Institute corresponds with practitioners, organisations, standards bodies, regulators, insurers and prospective participants who see the same gap. There is no application, enrolment or payment form; early interest registration opens shortly; enquiries are answered by e-mail.
Programme enquiries
Mandate Engineering — Foundation is In Development — Not Open for Enrolment. Enquiring does not place you on a waiting list and creates no obligation on either side.