I say no to more projects than I take on. With four fixed-price builds a year I'd have to. Working out how to say it well took me a lot longer than working out what to say it to.
The first thing I look at is shape. If there's a single person who can make decisions, a defined outcome and a deadline, I'm interested. If the email opens with wanting to chat about possibilities, I'm probably not. That isn't me being precious about it, it's just that a fixed price needs something specific enough to put a number on.
The second is how long it runs. I won't quote on anything past six weeks. Eight-week projects are usually two four-week projects wearing a trench coat. Three-month engagements are agency work and I'm not an agency, so that no is an honest one. I'm just not the right shape for the problem.
The third is whether there's something to build at the end. I don't take strategy or consulting work on its own. Discovery workshops, AI roadmaps, design sprints, audits, there are people who do those genuinely well and I'm happy to point you at them. I'm only useful when something gets shipped.
The fourth is procurement. If the first email comes with a vendor questionnaire and a panel, I'm out. This whole thing runs on one person and one client trusting each other, and once it turns into a procurement process that doesn't work. I'm faster and cheaper and better when the person buying is the person who'll use the thing.
The fifth is brand and marketing work. I don't build websites, brand identities or marketing sites for clients. I work inside your product, not on how it's presented. That one alone accounts for about a third of what comes in.
What I do say yes to is fairly narrow. One AI capability, shipped end to end, in your codebase, by me, at a fixed price, in four to six weeks. Everything else is a no, said quickly and warmly so we can both get on with our week.
The thing I'd pass on to anyone earlier in their career is to write your no's down before you need them. The hardest ones to say are the flattering ones, where the budget is bigger or the name is better known. It's much easier to decide the rules when nothing is at stake and then apply them when something is.