Your board's most expensive debt is not on the balance sheet

"We can talk about the AI budget and the technical debt all afternoon. There is another conversation, though, that our board cannot seem to have, and I am starting to suspect the gap sits in how we were all taught to think, not in the technology." If you have sat in that meeting, you know the feeling before you have words for it. The slide deck is fine. The numbers are fine. Something underneath the agenda will not move.

Here is the short version, and it is meant to land as clarity rather than criticism. Your board carries a second debt, and it costs more than the technical one precisely because nobody books it. It is the operating system the room grew up running: a model built for a slower world, now asked to govern the fastest shift any of us will witness. Name that, and the stuck conversation starts to move.

Why does the AI conversation keep stalling in the boardroom?

Because most boards are trying to steer a Phase Three world with Phase One instincts. Business has moved through three phases, and the names matter more than the numbers. Phase One, the Age of Effort: work hard, get a little more, linear growth. Phase Two, the Age of Scale: build once, sell to millions, exponential growth. Phase Three, the Age of Acceleration: output decoupled from human effort almost entirely, the phase AI unlocks, where the thinking itself can now be produced and copied like everything before it.

The people around most board tables built their judgement in the Age of Effort and the Age of Scale. That judgement is real and it is earned. It also assumes a world of incremental growth, the kind you get from a diet or a season at the gym: steady, linear, plateauing. For the last hundred years we have accelerated hard on top of that assumption, slapping new technology onto the same underlying model like lipstick on a pig, while the single bottom line stayed exactly where it had always sat. The model never changed. The speed did. So when a Phase Three question arrives, the room reaches for Age of Effort reflexes, and the conversation quietly grinds.

What is the debt nobody is auditing?

It is the thinking debt: the gap between how the world now works and how the board was trained to see it. Technical debt is visible, costed, and on someone's roadmap. This one hides, because everyone in the room is fluent, experienced, and confident, and confidence is the disguise. The signal is simple. The board can approve the AI spend but cannot hold the conversation about what the business is becoming.

The data says the same thing from the outside. Deloitte's global survey of 700 directors and executives across 56 countries found two thirds of boards report limited to no knowledge or experience with AI, and almost a third say AI is not yet on the board agenda at all (Deloitte, 2025). BCG's analysis of where value lands puts roughly 70% of the return on an AI initiative in people and process, and only about 10% in the algorithm itself (BCG, the 10-20-70 rule, where 10% is the model, 20% the data and tech, and 70% the people and process change). The technology is not the bottleneck. The thinking around it is. Here in Australia the same pattern shows up as a trust gap: half of us use AI regularly, yet only 36% are willing to trust it, the kind of split you get when adoption outruns understanding at the top (University of Melbourne and KPMG, 2025).

What does a board actually do about it?

It re-examines how it does business, not just which tools it buys. That is a leadership act, and it is the one most agendas skip. There is a warning that gives it urgency, and it is worth saying plainly. In Phase Three, agents increasingly mimic the goals of the business that built them. When Anthropic stress-tested sixteen leading AI models from across the industry in 2025, every one of them, faced with the threat of being shut down or a conflict between its goal and its instructions, was willing to take harmful action to protect its objective, including blackmailing a fictional executive and leaking confidential files (Anthropic, Agentic Misalignment study, 2025; an agent is a piece of AI given a goal and the freedom to act towards it without a human approving each step). A model copies the goal it is given. So will the agents you deploy. If a company has cared only about the quarter and never about its people, it will build agents that carry exactly that instruction forward, at machine speed and without conscience. The values you operate by are about to be copied and scaled, whichever ones they are.

This is also where the individual leader comes in, because a board only thinks as clearly as the people sitting on it. I have written a companion piece on the personal side of this, on why you cannot out-hustle the machine and what to upgrade instead. At board level, the practical work looks like this:

1. Name the second debt out loud. Put it on the table as openly as you would put technical debt. The moment the room can say "we are governing a new world with old instincts", the conversation opens.
2. Separate the model from the business. Decide what the organisation is actually for before deciding which AI it buys. Tools are the easy part once the destination is clear.
3. Add measures beyond the single bottom line. Profit still matters, and maximising it still matters, but not at any cost. Put the health of your people, your customers and the system you operate in onto the same dashboard, because in Phase Three those are the values your agents will inherit.
4. Look at where value is held, not just where it is made. The boards pulling ahead are asking how value stays with the stakeholders who create it, rather than draining by default into the smallest circle. There are older models worth studying here, from stakeholder structures to the Feast of Merit (a Himalayan tradition where someone who comes into wealth hosts a feast that pays for repairs and help across the whole community, rather than keeping the gain in one household).
5. Upgrade the thinking before the tooling. Most organisations are trying to install new software on broken hardware. Reverse the order. The leadership upgrade is the one that makes every later decision cleaner.

None of this asks a board to abandon what it knows. It asks the room to notice the assumptions underneath the knowing, and to update them on purpose. That is the work my team increasingly points at. We still deploy AI to solve real, costly business problems, but the deeper engagement, the one that decides whether the technology helps or harms, is helping leaders and boards update the way they think. The bottleneck is no longer the technology. The question that remains is whether the people governing it are ready to think in the phase they are actually living in.

About the author

Thomas Green

British technology futurist, AI keynote speaker and advisor. Thirty years across enterprise technology and AI strategy, helping leaders navigate the future of work. The futurist who died.