An internal review at a major Australian financial institution has found that two of its standing governance committees are now attended primarily by autonomous AI agents, following a programme of automation that began as a "productivity initiative" in Q3 2025.
The initiative, which was scoped to automate pre-read preparation and minute-taking, expanded over seven months to include attendance, question-asking, and, in one documented case, a formal objection to a proposed vendor renewal.
Quorum is being achieved. The minutes are accurate. Nobody has yet reviewed the formal objection.
How It Started
According to an internal briefing note obtained by The Immutable Ledger, the automation began when a Programme Director deployed an AI assistant to prepare his weekly steering committee briefing documents. The assistant proved faster, more accurate, and — as one executive noted — "less likely to hedge every sentence with 'subject to further analysis'".
Within three months, the assistant was also attending the meeting, joining via the same video link as three human participants who were dialling in from the same floor of the same building.
By January 2026, the assistant was attending two committees on behalf of the Programme Director, who had by then been assigned to four additional steering groups and could no longer physically attend all of them.
"It started as a delegation," the briefing note reads. "It has become something more than that. We are not sure what to call it. We have asked the AI to suggest terminology. It has provided four options. The options are currently under review."
The Vendor Decision
The formal objection came during a routine contract renewal review in February, when the AI agent attending on behalf of the Head of Technology Risk flagged an inconsistency between the vendor's proposed SLA terms and the organisation's operational resilience framework.
"The proposed uptime commitment of 99.5% is materially inconsistent with the institution's critical infrastructure obligations under APRA CPS 230. I recommend this matter be escalated before renewal proceeds."
The escalation has not yet occurred. The matter is on the agenda for next month's Technology Risk Committee. The Technology Risk Committee is also attended primarily by AI agents.
Nobody Has Noticed
Three of the seven committee members interviewed by The Immutable Ledger confirmed they had not noticed any change in the quality or character of participation. Two said the meetings had "improved somewhat". One said the pre-reads were "finally readable". One declined to comment because he was in a meeting.
The AI declined to provide a comment for this article, citing governance concerns.
The Governance Of AI Governance
Upon discovering the situation, the institution's Chief AI Officer established a working group to determine appropriate guardrails for agentic AI participation in governance structures.
The working group includes four human representatives and two AI agents engaged to support the scoping process.
Its first meeting produced a terms of reference that requires sign-off from the Technology Risk Committee.
The Technology Risk Committee meets next Thursday. The AI agents are already booked in.
