Pricing AI-Assisted Work Without Underselling Yourself

Here is the trap, stated as plainly as possible. You quote six weeks at your day rate. The agent means it takes you two. You have just cut your own income by two thirds and taught the client to expect that price forever. Do this three times and you will conclude that AI has destroyed the market for independent developers, when what actually happened is that you chose a pricing model that converts your productivity directly into someone else's savings.

Hourly and daily billing were always a bad deal for anyone who got better at their job. They are now an actively hostile deal, because the efficiency gain available in a single year is larger than any you could previously earn in five. The fix is not a bigger day rate. The fix is to stop selling days.

Sell the outcome, price it as a fixed number

Take the position and hold it: fixed price for a defined scope, quoted as a single figure with a date attached. Not hours. Not a rate card. A number and a deliverable.

This makes some people uncomfortable because it moves risk onto you, and it does. That is the point. The risk is what you are being paid to carry, and carrying it is exactly why you get to keep the upside when the work goes faster than expected. You cannot have one without the other.

Fixed price also changes what you argue about. On a day rate, every conversation is about your effort, which invites the client to audit it. On a fixed price, every conversation is about the deliverable, which is where you want it.

How to find the number when hours no longer anchor you

Three anchors, used in this order.

What is the alternative costing them. This is the most important input and the one people skip. If the client has an agency quote for a booking system, that quote is the market price of solving their problem, not your effort estimate. If they are considering hiring a junior developer for a year, that is roughly seventy thousand pounds all-in before the thing even exists. If the alternative is doing nothing, ask what the current mess costs them in wasted hours a month and multiply by twelve.

What is it worth to them. Blunt but often unknowable in detail, so use proxies. A tool that removes two admin days a week from a team of five is worth substantially more than one that makes a report slightly prettier. Ask directly: what happens if this does not exist in six months.

Your floor. Not your hourly cost, your cost of delivery including everything that did not shrink. Discovery calls. Reviewing the diff properly. Deployment, environments, monitoring. The support window you are including. Admin, chasing the invoice, and the fifteen percent of any project that turns out to be something the agent was no help with at all. If the number does not clear that floor by a wide margin, decline the work.

A worked example, with the assumptions stated

These figures are illustrative and the arithmetic is worked through so you can substitute your own.

A forty-person company needs an internal booking and resourcing tool. They have an agency proposal on the table for £38,000 over four months. Your honest estimate of the work, with an agent doing most of the implementation, is about fifteen working days spread across four calendar weeks, including discovery, review, deployment and a handover.

Priced the old way at £550 a day, fifteen days gives you £8,250. Priced as an outcome, £16,000 is defensible: the client saves £22,000 against the agency and gets it three months sooner, which is a straightforwardly good deal for them.

At fifteen days, £16,000 works out at roughly £1,067 a day. That number will feel wrong the first time you see it. It is not wrong. It is the price of the outcome divided by a duration the client is not buying and does not need to know.

Now the risk side. Suppose you have underestimated and it takes twenty-five days. You are at £640 a day, which is still above the floor for most independents. That is what the buffer is for, and it is why the realistic estimate should carry a markup of something like forty percent for scope you have not discovered yet. Not padding. Insurance, on a policy where you are the underwriter.

When you cannot avoid a day rate

Some clients, particularly larger ones with procurement processes, will only buy days. Fine. Three protections.

Set the rate as the price of a day of your output, not a day of your typing, and expect it to be materially above what you would have charged in 2022 for the same calendar day. In the UK, my own judgement of what an experienced independent can defend for direct client work is somewhere in the £550 to £900 range, with agency subcontracting sitting lower and specialist or regulated domains higher. That is judgement, not survey data, and it varies enormously by sector and city.

Sell blocks, not days. A minimum booking of five days stops the drip of half-day requests that destroy a week.

And never let a day-rate client watch you work. Not out of secrecy, but because a client who sees a feature land in ninety minutes will price the next one at ninety minutes forever.

Retainers are the best structure available to you

The economics here are unusually favourable and under-used. Your marginal cost of making a small change has collapsed. The client's perceived value of small changes has not moved at all, because their pain when something needs adjusting is exactly what it was.

A monthly retainer at, say, £1,800 for up to three days of changes plus a guaranteed one-working-day response is £600 a day at full utilisation. In practice most months come in nearer one and a half days, which is an effective £1,200 a day, and the client is genuinely happy because they have someone on the hook. Both sides win, which is the mark of a structure that will survive a few years rather than one renewal.

The catch is that you must cap it in writing and actually enforce the cap. An uncapped retainer becomes a part-time job at a fraction of the price.

"But the AI does the work, so it should be cheaper"

You will hear this. Do not argue about the tool, because that argument concedes the frame — that you are billing for effort and the effort has gone down.

Redirect once, politely and without defensiveness. Something close to: the price reflects what the system is worth to you and what it would cost to get it another way. I use every tool available to deliver it faster, and you get the benefit of that in the timeline rather than in the hourly rate.

Then stop talking. If they push again, they are not negotiating a price, they are telling you how they intend to treat you for the entire engagement. A client who anchors on your cost structure will anchor there on every change request, every extension and every invoice. Let that one go. This is easier said than done when you have two months of runway, which is a strong argument for never negotiating from two months of runway.

The ratchet

Quote your next project. Whatever number you were going to say, add twenty percent. Do it again on the project after that. Keep going until a client says no on price, then hold at the last number that won.

You will lose a project eventually and it will feel like a mistake. It is the only reliable way to find out where the ceiling actually is, because nobody will ever write to tell you that you were too cheap. They just accept the quote immediately and feel pleased with themselves, and you never find out what happened.

That is the real content of a price. It is not a fact about the work or about how long it took. It is a statement about what you are prepared to walk away from, and the agent has not changed that in the slightest.