AI made me faster. My experience is still part of the price.

“But if you’re using AI, shouldn’t it be cheaper?”

If we’re working by the hour and AI cuts the hours needed, yes: you get a smaller invoice. My hourly rate reflects my experience. For a fixed-price project, we agree on a delivery and a price. Faster execution lets you get that delivery sooner.

I have worked in this industry for 25 years. With LLMs and AI, a lot of the work now goes ridiculously fast. I can move between design and development and check an implementation while I have the whole idea in my head. Work that used to involve a lot of repetitive typing barely interrupts me anymore.

Those 25 years shape how I use the tools. They are part of why it goes fast.

The tool does not know what I should accept

I usually have a clear idea of what I want before I ask a model to build anything. I know where a seemingly small request can turn into a mess and which assumptions I need to check first.

I have had to fix enough of my own work after shipping it to recognise a wrong approach earlier next time.

A model can produce a convincing implementation of a bad idea very quickly. I still have to notice when a layout breaks with the content it needs to hold, or when a feature adds maintenance without helping anyone. Sometimes the most useful thing I do is delete half of what the model just wrote.

And when I deliver something, I am responsible for it. “The model wrote that part” is a useless answer to someone whose website is broken.

Speed gives me more room to get it right

A first version can run quickly. Getting it ready to hand over takes more work. Clear designs and finished content help enormously. If we’re figuring out the structure or how something should behave, we need to make those decisions before speed helps much.

Then I need to check what happens on a smaller screen, with a longer title or when there is no content yet. The person who inherits the work should be able to edit it without calling me every time.

AI helps with plenty of that too. A model can investigate a bug while I work through a design decision. I have more capacity to cover a project myself, including time to check what the model found and whether the fix holds.

If I can deliver earlier, great. There is no reason to keep it sitting around to make an estimate look respectable. I would rather use the time we gain to improve what we are delivering.

What is the website worth to the business?

Say a business generates US$1 million in annual revenue. A new website could help it reach US$2 million by making the product easier to find and buy. That is a hypothetical example, but it is the kind of conversation I want to have.

If that potential is credible, the price should reflect the commercial value of the work. The number of hours I spend writing code tells you very little about that value.

I would want to understand what is holding sales back and how we will measure the website’s contribution. Doubling revenue does not automatically double profit.

And I cannot take credit for every extra sale. The product, marketing and sales team all contribute. Any price tied to an expected outcome needs an honest view of what I can influence and the risk each side is taking.

Once we understand those things, we can agree on the delivery and a price that makes sense for both of us. If my experience and my tools let me get there faster, the business gets that value sooner.

I pay for the tools too

I pay for LLM subscriptions, API calls and the electricity to run models locally. Those costs come out of my business. They need to be covered by what I charge.

The customer is buying my craft, my judgement and my responsibility for the delivery. I invest in tools to do that work better. That investment does not automatically entitle someone to a discount. I can cover the costs in my price without putting every API call on the invoice.

With a fixed price, faster delivery can improve my margin even after those costs. The customer gets the agreed result sooner, and I get a return on investing in my own workflow.

I always ask for the budget now

Before I put together a proposal, I now always ask: what budget do you have for this?

Getting an answer is surprisingly difficult. When the spending was planned in Q4 of the previous year, there is already a number to work with. Keeping it hidden and then saying the proposal is too expensive leaves me guessing at a constraint we could have discussed from the start.

The budget changes what I should propose. With less money, we might keep more of the existing website or do the work in stages. I can work with a smaller budget. I need to know what it is, and I should be able to explain how the scope and price fit it.

I am happy that the tools have caught up with how quickly I want to work. I want that speed to help me deliver something useful for your business, at a price we both understand.