Who Decides? AI, Humans and the Future of Asset Management

Where should AI end, and human judgement begin, in asset management?

A community-led AMCouncil inquiry. Over five webinars, the asset-management community decides — decision by decision — what AI should do, what humans must keep, and what evidence is needed before we trust the outcome.

Next: Webinar 5 — Synthesis — the Boundary Framework v1.0 · date to be confirmedSee what the room said

4/5

Webinars complete

Living outputs

1577

Boundary votes

2798

Responses captured

The Human–AI Boundary Framework

The boundary between AI and human judgement isn't set by the technology — it's a governance choice, drawn decision by decision across the asset lifecycle.

We practise what we’re testing

This inquiry is run the way it argues for. AI does the heavy lifting — gathering research, drafting and summarising the room — but the judgement stays human. Volunteers from the asset-management community hold the boundary: theydecide what’s authoritative, what gets adopted, and where AI’s role ends.

It's not what AI can do

It's what AI should do at each decision — knowing the asset, judging its condition, planning the spend, directing the work — and what humans must keep.

The boundary moves with the stakes

Cleaning a data field and authorising a field intervention are not the same call. How far AI goes should rise with the risk and the consequences.

The community draws the line

Across five workshops, your votes, cases and disagreements become Version 1.0 of the framework — seven living outputs.

W1✓ Done

Knowing the Asset

✓ Done · 25 June 2026

W2✓ Done

Condition & Risk

✓ Done · 1 July 2026

W3✓ Done

Planning & Investment

✓ Done · 15 July 2026

W4✓ Done

Work & Operations

✓ Done · 30 July 2026

W5Next

Synthesis — the Boundary Framework v1.0

The seven living outputs

What this inquiry produces. Each publishes here, read-only, once a named human signs it off — nothing is official before then.

Decision Rights Matrix

In progress

Per asset decision: the AI-role band and the human rights that must be retained.

inputs: W1–W3W3 — the boundary vote — 72% capped AI at analyse-or-recommend, 0% autonomous; 56% keep owning the value judgement (n=18)

Boundary Framework

In progress

The principles that govern where AI ends and human judgement begins.

inputs: W1–W3W3 — three principles rated live — the deferral ledger 4.07/5, the signed objective 3.73/5, the frontier right 3.39/5

Lifecycle AI Map

In progress

Where on the asset lifecycle AI is helping, and how far it should go at each stage.

inputs: W1–W3W3 — the 'Planning & investment' stage mapped

Boundary Questions Toolkit

In progress

The questions to ask before trusting an AI output enough to act on it.

inputs: W1–W3W3 — where are your value weights written down? — only 39% could have a named person show them today (n=36)

Failure Modes Library

In progress

Named ways AI quietly goes wrong on asset data — so they can be designed out.

inputs: W1–W3W3 — the vendor-default weights nobody chose — 85% would not have signed them (n=39)

Open Questions Register

In progress

The unresolved tensions the community is still arguing about.

inputs: W1–W3W3 — Sim 3 split three ways on every plan; only 32% had a named owner for checking the optimiser (n=41)

Recommendations

Not started

What AMCouncil will ask of members, vendors and regulators. Drafted at W5.

—

Share a case

Tell us where AI helped — or could have hurt — how you know your assets. Members can add real examples to the case library; every submission is reviewed before it’s published.

🔒 AMCouncil members can submit a case to the library. Sign in or join AMCouncil →