Board readiness for an agentic workforce is a board’s capacity to govern an organisation in which AI agents carry out a growing share of the work, and may eventually outnumber the people who supervise them. It covers who sits on the board, what the board sees, how often it looks, and who in management answers for work that no employee performed.
The idea travels under several names: agentic workforce governance, digital workforce oversight, human-agent workforce governance, board oversight of agentic AI and, more loosely, AI-era leadership. Each describes the same shift. Oversight built for a workforce counted in people now has to cover a workforce that is partly software, and that software can act as well as advise.
The frameworks boards already rely on point the same way. ISO/IEC 42001:2023 requires top management to demonstrate leadership of the AI management system and to assign roles, responsibilities and authorities for AI. The NIST AI Risk Management Framework (AI RMF 1.0) places accountability structures and senior leadership responsibility in its Govern function. The UK Corporate Governance Code 2024 asks boards to monitor and review the effectiveness of the risk management and internal control framework, and, for financial years from January 2026, to declare whether material internal controls were effective. Controls that an agent performs are still controls. Under the EU AI Act (Regulation (EU) 2024/1689), deployers of AI systems must take measures to ensure sufficient AI literacy among the staff who operate them.
Most boards are not yet equipped for this. Two-thirds of board members and executives (66%) say their boards still have limited to no knowledge or experience of AI, and 40% say AI has caused them to think differently about their board’s make-up (Source: Deloitte, Governance of AI: A critical imperative for today’s boards, 2nd edition, 2025).
This analysis draws on a recent conversation with Jeff Epstein, operating partner at Bessemer Venture Partners, former chief financial officer of Oracle and a director of AvePoint. Part 1 of the conversation, on the finance function, is covered in The CFO’s AI Mandate.
Why does a workforce of AI agents change what the board oversees?
It changes oversight because the board’s instruments assume that work is done by people who can be counted, managed and held to account. Headcount reports, organisation charts, spans of control and delegations of authority all describe human employees. The premise of the discussion is that AI agents will soon outnumber your team. When that happens, a growing share of the organisation’s decisions, transactions and customer contacts will sit outside every report the board currently reads.
The risk is less dramatic failure than quiet multiplication: each agent approved as a small productivity gain, until nobody can say how many there are, what they may do or who answers for them.
What makes a board effective when the future is this uncertain?
An effective board spends its time on decisions that have not yet been made, rather than on reports of what has already happened. Epstein has made this argument elsewhere, in another interview. In a recent conversation he extends it to leading through uncertainty, drawing on Charlie Munger’s mental models as a way to reason clearly when the facts are incomplete.
For AI, the implication is direct. The decisions that matter, such as how much autonomy to grant, which processes to hand to agents and how to redeploy people, are all decisions not yet made. A board that only receives updates on AI projects already under way has ceded those choices to whoever moves first.
The AIPEX view: govern agents as a workforce, not as software
At AIPEX we think boards should stop treating AI agents as an IT asset and start treating them as a workforce, with the same instruments the board already uses for people. Every agent should have a mandate, a manager and a place in a register the board can read.
| Human workforce instrument | Agent workforce equivalent | Board question |
|---|---|---|
| Headcount report | Agent register: number, purpose, owner | How many agents act on our behalf today? |
| Job description | Agent mandate: tasks, data, limits | What may each agent decide alone? |
| Line manager | Named accountable executive | Who answers when it gets something wrong? |
| Delegation of authority | Autonomy thresholds and approvals | Where must a person still sign off? |
| Performance review | Monitoring, error rates, incidents | Is the agent doing what we approved? |
The mandate itself is the subject of Who Owns the AI Agent When It Breaks?, which sets out how to write delegated authority for an agent before it reaches production.
How should board composition change, and what does it require from the chair?
Board composition should change so that at least some directors can challenge AI proposals on their merits, not so that the board acquires a single technology expert. The better aim is a baseline of AI literacy across the whole board, with deeper expertise held by one or two members or brought in through advisers.
From the chair it requires three things: a skills matrix that names AI literacy explicitly, a development plan for directors who lack it, and agenda time that is protected for forward-looking AI decisions. Literacy grows fastest through use, which is why we argue that leaders should prompt first.
How often should the board review AI agents?
The board should review its agent workforce at least quarterly, with the register, material incidents and any change in autonomy reported as a standing item. Changes that increase what agents may do without human approval, or that move them into customer-facing or financial decisions, should come to the board or a delegated committee before they take effect, not after.
What should the board ask management now?
- How many AI agents act on our behalf today, and who keeps the register?
- For each agent that can act rather than advise, who is the accountable executive, by name?
- Which decisions have we delegated to agents, and where is that delegation written down?
- Does our skills matrix name AI literacy, and how many directors meet it?
- What share of our board time is spent on AI decisions not yet made, rather than reports on work already done?
- How will our workforce plan change if agents outnumber staff in key functions within three years?
If nobody at executive level can answer these, the missing piece is usually ownership. A Fractional Chief AI Officer can hold the agent register and bring agent risk to the board as a standing item, without a full-time hire. The Executive Strategy Day works through the decisions not yet made: which work to hand to agents, on what mandate and under whose authority.
Frequently asked questions
What does it mean for a board to be ready for an agentic workforce?
It means the board can govern an organisation in which AI agents perform a growing share of the work. That requires the right skills around the table, regular visibility of what agents do, and a named executive accountable for each agent that can act. For a full explanation, see When AI Agents Outnumber Your Teams: What Boards Must Prepare For →
Should a board appoint an AI expert as a director?
A single specialist director is rarely the answer, because the rest of the board tends to defer to them. A better aim is baseline AI literacy across the whole board, with deeper expertise held by one or two directors or brought in through advisers. For a full explanation, see When AI Agents Outnumber Your Teams: What Boards Must Prepare For →
Who is accountable when an AI agent makes a wrong decision?
A named executive should be accountable for every agent that can act rather than advise, under a written mandate that sets out what it may decide alone and what needs human approval. Without that mandate, accountability defaults to nobody. For a full explanation, see Who Owns the AI Agent When It Breaks? Delegated Authority for Boards →