Article
The Last Human Job
AI can assess risks and run a business. The people who govern it still set priorities, honor commitments, and answer for the consequences.

Suppose an AI could run my business. It could find customers, negotiate contracts, schedule the work, and collect payment. One morning it tells me we can take on a much larger client. The work is profitable. There’s one catch. If the client pays late, we could miss payroll.
I ask how likely that is. I ask what cash reserve would cover us, what it would cost to borrow, and whether we can negotiate a deposit. Suppose I get excellent answers to every question. I still have a decision to make. Am I willing to put someone’s pay at risk so I can grow my business?
My answer would be to ask for the deposit, even if it costs us the contract. I made a commitment to the people who work here. The larger client is an opportunity. When the two compete, I know which comes first.
I could give the AI that policy in advance, and it could make the call without waking me. That would be good delegation. It would also tell you something about how I govern the business.
Where the priorities come from
Risk management comes close to describing what remains for people as machines take on more of the work. Almost every consequential choice means accepting one risk to reduce another. Decline the contract and I could lose a customer I’ll need next year. Accept it on the client’s terms and my employees bear some of the cost if my forecast is wrong.
An AI can help me see both risks. It can even challenge my priorities. Perhaps I’ve been so cautious about taking on work that I’ve put the business, and those same jobs, in danger. I would want to hear that argument. But a recommendation to grow faster depends on a view of what the business is for, what it owes people, and which risks it may reasonably ask them to bear. Those are questions of governance. Someone with authority over the business has to answer them.
In The Judge, I wrote about the person who directs an inquiry, weighs the record, and answers for a decision. Governance also requires judgment, but the person exercising it stands in a different relationship to the matter. A judge hearing a dispute over a contract owes both parties an impartial decision. A board considering whether to sign that same contract has charge of the company’s affairs. Its members ought to care how the company fares. If a deal would benefit a director personally, that director has particular reason to ask whose interests the deal would serve.
I govern my own affairs when I decide what commitments to make. A board exercises that authority collectively, on behalf of the entity entrusted to it. In either setting, governance includes setting priorities, choosing who may act, and staying able to intervene when the assumptions behind a decision change.
What remains when the work is done
I don’t know which operational task will be the last to pass to a machine. Perhaps it will involve delicate physical work or caring for someone in distress. My argument holds even if machines eventually do all of it well.
Imagine the business runs for a year without my having to approve a single contract. Then an employee asks why we passed up work that could have kept her team busy. She deserves an answer. I should be able to explain the commitment behind the deposit rule, hear what the rule cost her team, and reconsider it in light of what happened. If all I can tell her is that the system declined the work, I’ve lost the practical ability to govern a business I still claim to run.
That is the sense in which I think governance is the last human job. Job is an imperfect word. Nobody hires me to govern my own affairs, and a board need never perform the work it authorizes. What remains is a responsibility. However capable the people or machines acting for us become, we still owe those affected by our decisions an account of what we asked for, why it mattered, and what we will do when we learn we were wrong.