We’re a small software company based in Amsterdam. And we’ve been using AI coding agents big time.

Recently we’ve been figuring out how to use them in our daily work, which mostly entails doing tickets. 

Now, coding agents aren’t much good at doing ordinary grunt work. They’d much rather build something new, a characteristic they have in common with human developers. 

But we have built a system for them to do tickets efficiently and relatively autonomously.

Recently we’re having our coding agents log our time, and to that they’re told to add the estimate for a human developer, which they do quite well. 

Which gives me the dev time x 3 estimate.

AI coding agents make us at least three times more productive, sometimes much more. Should we then also charge our clients three times more? No.

That would amount to tripling our hourly rates.

But we have a better idea. 

We’re going to stop calculating time & materials altogether.

Doesn’t mean that our clients will get our time for free. They’ll get it at a discount, in return for commitment. 

We’re offering long term contracts against a fixed monthly fee.

How we calculate a fixed monthly fee is a combination of averaging, complexity algorithms, and good old-fashioned estimating.

For existing clients, that’s relatively easy. We calculate the average yearly expenditure across the last five years, subtract larger projects, leaving costs of support, maintenance and small changes. Divide by twelve, add a discount, and that’s the monthly rate.

For clients coming in with existing projects, we have a different approach. 

In the software world, there are tried and tested methods of defining complexity. 

Even before AI, we were thinking of ways to improve our services while removing the constant discussions about time spent on tasks. In that context we’d explored the research into software complexity and built a prototype. 

There are two approaches that have proved viable, McCabe’s Cyclomatic Complexity, which quantifies the number of independent paths through source code and Halstead’s metrics which analyse operators, operands, and program vocabulary to estimate complexity. We investigated both, and created a hybrid formula which works quite well. 

So when a client asks us to take over an existing project, we’ll set the monthly subscription by applying our estimation method. 

What about new projects? 

We feel we have a truly unique offering for new projects.

Most aspiring businesses need to upfront a large amount when building their platform. They could get investors interested, but that comes with ownership strings attached. 

Instead, they could ask us. Investments of up to a certain amount can be converted into a monthly subscription tied to a long term contract. 

To give an example, imagine you’re a small business that has an idea for a unique digital product that warrants a custom software solution. The estimated cost is a hundred thousand euros. You might have vibe coded a prototype, or otherwise demonstrated the estimate holds water.

Depending on your schedule, our company would offer a two year contract which could either be a fixed monthly fee, or one with a bump at the beginning if you’re anxious to bring your product to market quickly. Either way, you’ll be smearing costs out across two years instead of having to hand out a huge sum upfront.

Businesses like us can help businesses like you. We can help each other.

We think it’s only fair we share the productivity gains AI brings with our clients and with aspiring new businesses. And we think we’ve found a way to make it good business sense for both of us. 

In the end, doing business isn’t a zero sum game. As has been proven since times immemorial, businesses exist and thrive next to each other. 

We aim to follow that tradition. 

Let us know what you think.

Header image: Metered Man, woodcut, 17th century.