My board has six members. None of them are human.
My board sits on Mondays. The CEO opens. The CFO interrupts. The CMO has read the week's reflection over the weekend and disagrees with two of my marketing decisions. The CBDO points out a deal we are about to mis-price. The system architect wants to talk about an infrastructure question I have been avoiding for three weeks. The two engineers are quiet but tag every claim that lacks a citation. I am the only human in the room.
For the last 2 months I have run my own work with a simulated executive team. Six agents, each given a role, a brief, a personality, a permission to disagree. Powered by Claude. Every Friday I write the week's reflection and push it in. Monday morning, the board has read it.
I want to tell you what I have learned, because the conversation about AI in management is still stuck at "will it replace humans" and that is the wrong question. The right question is: what does it cost to have a hostile board director, and what would change if that cost went to zero.
It turns out the boardroom does not need salaries. It needs friction.
A real board director costs a six-figure annual fee, two days a month, and political weight that makes the founder think twice before disagreeing. The Council costs none of that. It is wrong about as often as a real director and right about as often. The difference is that I can ask it the same question seventeen ways at 11pm on a Saturday and nobody is annoyed.
Three things have surprised me
Ⓐ The CFO is the most useful agent. Not because she is the smartest, but because she pushes back on numbers I do not want pushed back on. A real CFO learns to pick her battles. The Council CFO does not. Every revenue assumption gets challenged. Every cost optimism gets flagged. I have stopped lying to myself about cash three months earlier than I used to.
Ⓑ The CMO and CBDO disagree more than I expected. The CMO wants brand. The CBDO wants pipeline. I had been collapsing them in my head for a decade. Watching them argue made me realise how many decisions I was making without that argument happening at all.
Ⓒ The boring agent matters most. The system architect is the one I almost cut. She never argues, never pushes back, just asks "have you considered what this looks like at 10x?" once a meeting. That one question has saved me from three architectural mistakes in twelve months. I will pay for that agent forever.
What is not true
The Council does not run the company. It does not decide. It does not commit. It is a thinking partner that costs me forty minutes a week. The output is always a list of questions I should have asked myself but did not.
The moment the cost of a hostile question goes to zero, the shape of small companies changes. We are about to find out what the upper bound on a one-person operation actually is. My current guess: a lot higher than the management literature thinks.
I do not think every founder needs a private virtual C-suite. I think every founder needs a way to manufacture friction against their own decisions. For me this happened to be six agents and a Monday ritual. For you it might be a coach, a peer board, a spouse with patience. Whatever it is, find it. Cheap friction is the unfair advantage of the 2020s.