Status Droid how long software work actually takes

Hourly Billing and Unstable Speed

Hourly billing works when hours and value move together. Get better and you produce more per hour, so you raise your rate, and the arrangement stays fair.

That has always been slightly broken — it penalises expertise, since the person who solves something in twenty minutes bills less than the person who takes three hours. It became substantially more broken when the relationship between time and output stopped being stable within a single project. For the operational side of recording work and turning it into client-facing evidence, see productivity versus efficiency.

The two failures, both real

You get faster and earn less. Work that used to take eight hours now takes two on generation-heavy tasks. You have just cut your income by 75% on that work while delivering the same outcome. Nobody thanks you.

You get slower and look like you are padding. Verification-heavy work can now take longer than it used to. You bill more hours for the same deliverable, and the client — who has read that AI makes developers faster — draws an unflattering conclusion.

Both are honest. Both damage you. And which one you get depends on the task, not on you, which is what makes it impossible to price away with a rate change.

Why the usual fix does not fix it

"Raise your hourly rate to compensate" assumes the speedup is uniform. It is not. Raise the rate and you overcharge for verification-heavy work while still underearning on generation-heavy work. Upwork publishes practical material for freelancers and client work See Upwork Resources.

You have applied an average to a bimodal distribution, which is the same error as adding a flat percentage to an estimate.

When hourly still works

Not an argument to abandon it. Hourly remains the right instrument in three situations.

Genuinely open-ended work. Investigation, ongoing support, "help us with whatever comes up." Nobody can scope it, so nobody should pretend to.

When the client controls the direction. If they will change their mind repeatedly and that is fine, hourly prices that honestly and prevents a scope fight.

When trust is not yet established. A small hourly engagement is a low-risk way for both sides to find out whether this works.

For everything with a definable outcome, hourly is now pricing the wrong variable.

The alternatives, honestly

Fixed price per outcome. You carry the variance, which is the point — you are the one who can estimate it. Works when the scope is genuinely definable and you have enough data to price it. Fails if you cannot, and the data has to come from recording rather than recall.

Day rate. Underrated middle ground. Coarser than hourly, so small speedups do not reduce your income, and it preserves the simplicity clients like. The unit is a day of your attention, not an hour of typing, which is closer to what is actually being bought.

Retainer. A monthly amount for availability and an agreed scope of work. Best fit where the relationship is ongoing, and it removes the conversation entirely — nobody asks whether AI made this month cheaper.

Value-based. Fashionable and hard to execute. Requires the client to agree on the value beforehand, which most small clients will not do.

For most small studios the practical move is day rate or fixed price for defined work, retainer for ongoing, hourly kept for genuine unknowns — what each instrument actually prices decides which fits.

Making the switch without losing clients

Do not announce a philosophy. Nobody wants an essay about pricing models. Quote the next project the new way and see how it goes.

Quote a number they can compare. If your day rate lands where eight hours used to, the transition is invisible. Repricing and restructuring at the same time is two arguments.

Keep the estimate transparent. Showing which parts are generation and which are verification does more for acceptance than any pricing model, because it moves the conversation to the work.

Hold rates on verification-heavy work. This is where the discipline is required. It is the work that got harder, it is where the client's suspicion is highest, and discounting it teaches everyone the wrong lesson.

The uncomfortable part

Under fixed price you keep the upside when generation goes well. That feels like getting paid for what a model did.

The counter is that you also keep the downside when verification takes three times as long, and you are the one who knew which case this was. You are being paid for the judgement about which task was which, and that judgement is now worth more, not less — it is the scarce thing in this whole arrangement.

If that feels uncomfortable, price closer to your costs and keep the peace of mind. Just do it deliberately rather than by leaving an instrument in place that has stopped measuring what it used to — and there are four signals that the moment has arrived.

The short version