There's a defence of hourly billing that sounds reasonable right up until you actually run the numbers: "when nobody knows what the outcome is worth yet, time is the only fair thing left to charge for." I've made that argument myself. It's wrong, and it's worth working through why.

The Maths Nobody Runs

Take two developers. A junior, early in their career, and a senior with years behind them. Same brief, same outcome required. The junior takes 100 hours to get there. The senior, because they've solved this shape of problem before, gets there in 20.

Pay them both the same hourly rate and something perverse happens: the junior earns five times more for the same result. Pay the senior double the junior's rate to reflect their experience, and they're still earning less overall, for work that's probably better, because the fee is anchored to hours logged rather than the problem actually being solved.

Hourly billing doesn't just fail to reward skill. It actively penalises it. The faster and better someone gets, the less they're allowed to earn for the same outcome. That's not a pricing quirk, it's the model working exactly as designed, and the design is backwards.

Why "We Don't Know The Value Yet" Isn't An Excuse

The usual escape hatch is early-stage work: research, invention, anything where nobody can put a number on the outcome before it exists. Surely then hourly is fair, because there's nothing else to price?

There's an old story, probably more parable than history at this point, about an inventor closing in on a working lightbulb. He's already burned through thousands of failed attempts. Somewhere along the way, another man had quietly worked out how to make the filament last, not the whole invention, just that one piece. The inventor didn't want it at the time. He was funded by the hour, so grinding through the problem himself cost him nothing extra and paid him for every failed attempt.

Then someone else changes the deal. An investor turns up and says: I'll pay you for how fast you get there, not for how long it takes. Suddenly that filament trick is worth finding. It might save five thousand attempts. The inventor goes back, does the deal, and gets to a working prototype in a fraction of the time, because for the first time, speed itself had a price.

Nothing about the outcome's eventual value changed in that story. What changed was what was being paid for. You can price speed and certainty long before you can price the outcome. You just have to decide that's what you're doing, instead of defaulting to the clock because it's the easiest number to write on an invoice.

Where This Actually Bites

This isn't only a freelancer-pricing problem. Publicly funded research and development often runs on close variants of hourly pay, salary pro-rata, time logged, audited to make sure nobody's taking the mick. It sounds fair. It isn't, for the same reason the junior-versus-senior maths isn't fair: it makes cheaper, slower labour more profitable per pound spent than skilled, fast labour, on exactly the projects where getting to an answer sooner is the entire point.

AI sharpens this problem rather than solving it. When a skilled person uses AI to do in two hours what used to take two days, hourly billing punishes them for it, their invoice shrinks for getting better at their job. The incentive should run the other way: reward the outcome and the speed of getting there, and let the tools someone uses to do that be their business, the same way nobody asks a plumber to bill less because they own a better set of spanners.

The Actual Rule

If you can't price the outcome yet, price the thing you can see clearly: how much faster or better a specific piece of skill or insight gets you there. That's a real number. It's just rarely the one anyone bothers to ask for, because "time spent" is so much easier to write down than "value of not wasting the next six months."