Future professions

Writes machine-readable mandates: what agents may do, for whom, under what limits, with what evidence. A proposed title, conceptually rooted in EU AI Act Article 14, mechanism design and the delegation framings used by frontier labs.

Human: authoring authority. Machine: execution and logging.

Designs an organisation's delegation topology: which decisions go to whom, human or machine, with what escalation. Draws on secure delegation as a research theme at MIT; no degree teaches it.

Human: architecture. Machine: routine decisions.

Non-human identity, credential brokering, OAuth 2.1 and per-tool authorisation scopes for agents. Advertised by infrastructure and security employers; standards work under way at the IETF, the Cloud Security Alliance and NIST.

Human: identity design. Machine: authenticated action.

Signed policy bundles, runtime policy verifiers, trust roots and revocation for agent populations. A principal-engineer posting for exactly this role was live in 2026.

Human: policy authorship. Machine: enforcement.

Sets budgets, quotas and revocation policy across an agent population; treats human intervention as a costed, finite resource. Extrapolated from human-in-the-loop operations roles that already manage hundreds of contractors and multi-million-dollar budgets.

Human: economics. Machine: work.

Designs the machine-generated evidence trail an auditor or regulator will read. Implied by evidence-first professional methods and by the measure-and-manage functions of the NIST AI Risk Management Framework.

Human: evidence design. Machine: emission.

Golden datasets, expert rubrics, model-as-judge autoraters. Advertised by evaluation vendors and frontier labs.

Human: measurement design. Machine: scoring at scale.

Designs organisations composed of human roles, persistent agent roles and policies. The closest existing analogues are an unaccredited executive programme and consulting frameworks for the agentic organisation.

Human: organisational design. Machine: composed operation.

Prices errors-and-omissions and liability exposure of enterprise AI deployments. A pre-launch specialist insurer was recruiting a senior underwriter for this in September 2026, seeking 8–12 years of experience for a role no degree trains for.

Human: risk selection. Machine: portfolio monitoring.

Measures autonomy, artefact value and wage displacement from usage telemetry. Practised inside frontier-lab economic-index teams and labour-economics groups.

Human: econometric method. Machine: telemetry.

Designs hierarchical automated supervision in which humans occupy only the highest escalation tier. Formalised in published research on tiered agentic oversight and scalable oversight; not yet a hiring line.

Human: escalation policy. Machine: routine oversight.

Standards, reputation and settlement between agents transacting on behalf of principals. Active in agent-to-agent protocol work and in central-bank and payments research on agentic commerce.

Human: protocol design. Machine: transactions.