# Attention *The answer that feels responsible* --- Most of a senior job is deciding where things go rather than doing them yourself: where the money goes, where your own hours go, and where your few load-bearing people spend their time. Three of those calls come up more than the rest. --- **Spend capital on the structure, not the magic.** Most of the AI cases I've been asked to back die on the two dullest questions: what is the actual problem, and how will we know it worked. The ones worth backing are rarely the exciting ones. A faster first draft of a document saves real minutes, but it does not move revenue or cost by itself, and it is easy to wave through as if it did. What pays is the boring structural work: reworking the workflow underneath, and getting the data and systems into a state the tools can use. An agent does not see a business the way an org chart does - it does not care who reports to whom, it sees the work and the data. In most mid-size businesses the real bill is codifying what the place already knows, and that is the one people flinch at while the trade press fills with claims that someone else got the magic for free. The numbers usually dip while you are doing it, too - data quality drops the moment you start looking at it honestly ([[Inverse response]]) - and holding your nerve through that is part of it. **Give away the work you are best at.** You reach a senior seat partly by being good at something, so there will always be work you could do better than whoever you hand it to. The books say delegate; in a mid-size business that is only half right, because sometimes the quickest thing is to do it yourself, and being too grand for the spreadsheet is its own kind of failure. But the work you are best at is the stuff that ends up stuck to you. Handing it over means watching it done worse than you would do it, and the reflex is to grab it back. I did exactly that with one piece of work and spent the better part of a year as the bottleneck on it. Do that twice and the person you gave it to has learned it was never really theirs. Better to let it run at eighty per cent under someone whose job it now is than at a hundred under the one person who cannot afford to hold it. **Protect your best people from the urgent.** The scarcest thing you allocate is not your own time. It is the attention of your few load-bearing people - and they are always the ones the next fire wants too. Left alone, the urgent wins every time, because the fire is loud and has someone standing behind it, while the important work is quiet and easy to keep putting off. So stand in front of those people and take the requests yourself: some turn out not to need them at all, and the rest can wait or go in a batch ([[Flow and bottlenecks]] has the mechanics). A key person's queue fills faster than they can clear it - problems take longer to fix than they took to arrive - so it only grows unless someone guards it on purpose. And guarding it means letting a real fire burn, with a name on it and someone asking why you did not help. It is only ever obvious afterwards: a run of weeks where the urgent got all the good hours, and the important thing has not moved at all. ---