The billable hour is dead - Long live the fixed price!

If your e-commerce agency became three times as efficient, do you think you'd get an invoice a third of the size?

Probably not, right? And I don't think that's where you should be spending your energy either. What's happening at your suppliers right now is bigger than a pricing question. AI is killing the billable hour as a way to measure and charge for work. My hypothesis is that the billable hour is dead for e-commerce development, and that you as a buyer have just as much to gain from it being replaced with something more sensible.

Your supplier is caught in a squeeze

To understand what you might see in your next quote, you first need to understand the situation your supplier or consultant is in right now:

On one side, the hours are disappearing or shrinking. A developer working with AI delivers the same feature and value in a fraction of the time. For an agency that lives on billed hours, revenue per engagement is shrinking, and not by a little.

On the other side, there's a new cost: AI tokens. Today it's often small compared to the hours that disappear. But I don't believe the token prices we see now are the ones we'll be paying in a couple of years. The big AI players like Anthropic, OpenAI and Google are burning enormous sums to win market share and customers today, much like Uber did in its early years. At some point, the AI investors will want their money back... And even if the price per token falls, consumption rises when agents are left to work on their own for hours at a time. So the unit price of a token can go down while the bill still goes up.

Fewer hours to bill and a cost nobody knows where it will land. That's the supplier's problem and dilemma, but it quickly becomes yours too.

Two suppliers, two different ways of handling it

I've talked to two agencies who have solved this in completely different ways. Both examples say something about what you as a customer might run into.

The first one has built an advanced system to measure AI consumption in projects and ongoing maintenance. All usage from the company's AI accounts is collected centrally, each project is linked to the right customer, and the data runs through two layers of price lists: the AI provider's cost and the consultancy's own margin. The customer gets an invoice with both hours and tokens, broken down by project. The reasoning is that the token is the unit AI works in, the same way the hour is the human unit.

It's impressive and transparent. But as a buyer, I'd ask myself what I'm actually buying. Here, the token becomes the new hour. You're still paying for the effort, not for the result. If an agent gets stuck and chews through a million tokens down the wrong path, it's your invoice that takes the hit. And how are you supposed to judge whether the consumption was reasonable?

The second agency works with fixed prices, but calculated as cost-plus: estimated hours x hourly rate, plus a bit of risk margin. AI shrinks the time estimate and therefore the price. That sounds good for the customer, but the consultant's sales effort doesn't get any shorter or more efficient. Quotes, free workshops and requirements discussions take as long as before, and now the consultant may spend as much unpaid time selling the project as delivering it. And on top of that comes the AI cost. Simply put, a tough spot.

Why it's your problem too

The second example looks like a win for you as a customer, but I think it's a trap. A supplier who is suffering in your engagement will sooner or later do something about it. Perhaps putting in more junior consultants, less time for quality assurance and testing, higher prices for maintenance, or less interest in your business in particular. The cost of all that unpaid sales time doesn't disappear either. Someone pays for it in the end, and often it's you as the customer, one way or another.

Our experience at Commerce Mind is that the best and most valuable e-commerce partnerships last for many years. For that, both parties have to profit from them. Pushing the price down to a level where the supplier can't break even is a short-term victory. A true Pyrrhic victory. (I've been waiting to use that expression in real life...)

The models you'll encounter

Agencies will try different ways out of the squeeze. Here's how it might look from your side of the table.

• Higher hourly rate What it means for you: You pay more per hour, but for fewer hours. Watch out for: Not being able to compare with last year's quote.

• Tokens as an add-on to the invoice What it means for you: Full visibility into the AI cost, like materials from a contractor. Watch out for: You carry the risk of the consultant's inefficient AI use.

• Fixed AI fee per project or contract What it means for you: Easy to budget. Watch out for: A lump sum that is rarely revised downward.

• The supplier uses your AI accounts, not their own What it means for you: You control cost, data and code, in theory. Watch out for: You also pay for the supplier's mistakes and learning curve.

• The supplier does nothing What it means for you: Cheap, for now. Watch out for: A supplier whose margin is quietly being eroded.

• Fixed price with AI built in What it means for you: You buy a result, and the supplier carries the production risk. Watch out for: Making sure the price follows if/when AI gets cheaper.

Most of the models above try to save the billable hour in some way. Only the last one leaves it behind.

Buy results, not hours nor tokens

My recommendation to you as a buyer is: a fixed price with the AI cost built in. Not as a separate line item, but as part of the supplier's production cost. An silly example from another industry: a bakery doesn't send you a separate invoice for the electricity used by the oven when you buy a bun.

A fixed price with AI built in gives you three things:

• You buy what you actually want – a delivery – instead of the effort behind it.

• The supplier carries the risk if an agent gets stuck.

• You don't have to review token reports you can't assess anyway.

But it requires you to accept one thing: the supplier should be allowed to make money from becoming more efficient. Otherwise the incentive to improve disappears. The reasonable approach is to share the gains in some way. I believe fixed prices will fall gradually as tools and ways of working mature. The price curve for fixed-price projects points downward until what you might call "Peak AI" – the day the AI companies have to start taking profits. Then it will level off, or turn.

That's why I would write into the contract how prices are reviewed, in both directions, for (longer) projects and for maintenance. If AI gets cheaper, you should notice it in the next contract period. If it gets more expensive, the supplier should be able to raise it without it turning into a conflict between you.

Then there's something you can do yourself. If your consulting supplier's unpaid sales time is growing, make it a little cheaper to sell to you. Clear requirements, a reasonable procurement process, fewer pitches and fewer rounds of free workshops. It saves time for both parties, and it shows in the price and in the well-being of your suppliers.

The questions I think you should ask

At Commerce Mind, we always sit on the customer's side of the table when suppliers are chosen. If I were an e-commerce retailer today, I would ask my agency these questions:

• How do you use AI in our projects and in maintenance?

• How do you price it, and what happens to our contract the day AI prices go up or down?

• How have your prices changed since you started using AI, and how do you see them in two years?

• Whose AI accounts are used, and where does our code and data end up?

• Are you breaking even on this engagement?

That last question may feel strange to ask as a buyer. But a supplier who can honestly answer yes to it is a supplier who will still be around in three years.

In summary: the billable hour was never the value. It was just the easiest way to charge for it.