For most of Koombea's history, we sold software the same way most development companies still do: by the hour. It worked. But there was always one uncomfortable part of the conversation.
A client would ask, "What is this actually going to cost me?" And the honest answer was: we can estimate it. That isn't the same as knowing.
If a project took longer than expected, the client paid more. If we found a faster way to build something, we billed less. The contract rewarded time spent, even though time was never what the client actually wanted.
They wanted the software.
Key Takeaways
- Hourly billing rewards time spent, and time was never the thing clients actually wanted to buy.
- An AI Pod commits to a scope, sized in story points, at a price you approve before work begins.
- On fixed-bid work, if we underestimate within the agreed scope, we absorb the difference, not you.
- AI-First delivery makes that risk practical: written specs, tests derived from them, and gated merges that assume mistakes will happen.
- Twelve common project shapes have published prices, and hourly Staff Augmentation is still here when you direct the work.
AI Made the Mismatch Impossible to Ignore
Over the last few years, AI changed how we work inside Koombea. Not just the coding: we now use it throughout the delivery process, from scoping and requirements through implementation, testing, review, and deployment.
We call this AI-First delivery. The first benefit was the obvious one: a smaller, more focused team could simply ship more.
Hourly billing turned that gain into a problem. Every time we got faster, we billed less for the same feature. A client who wants something shipped shouldn't have to care whether it takes us 20 hours or 40, but the hourly contract made everyone care.
So we changed the model.
We Stopped Selling the Meter
With an AI Pod, we agree on what we're going to deliver before the work begins. We break the scope into clear items, each sized in story points, and you see the point total and the price before approving the work. Then the Pod delivers against that commitment in two-week cycles. The pricing model explains how we size points. And when your roadmap is still moving, you can buy the same points as standing monthly capacity instead of a fixed bid.
On a fixed-bid project, if we underestimated how difficult something would be within the agreed scope, we absorb the difference. That sounds like a small contractual change, but it moves estimation risk from the client to us. It's only practical because AI-First delivery gives us enough of an advantage across the development lifecycle to take that risk intelligently.
Of Course, There's Another Question
Usually it comes shortly after we explain the model. "How do I know AI isn't just generating a lot of bad code faster?" It's a reasonable concern. Our answer isn't that AI gets everything right. We design the process assuming mistakes will happen.
Work begins with a specification that defines what something should do, including its edge cases and acceptance criteria. Tests come from that specification, and the work isn't complete until it passes them. Automated checks and reviews gate what gets merged, and engineers remain responsible for the software that ships.
The goal is not to maximize how much code AI produces. The goal is to use AI to make a disciplined engineering process faster. That distinction matters: a prototype that looks convincing after two days is very different from software that survives production.
We Also Started Publishing the Numbers
Another thing has bothered us for years. Why should someone sit through a sales call just to find out whether they can afford the project? So we've started doing the opposite. We looked at the kinds of projects we build repeatedly, grouped them into common shapes, and published what twelve of them cost in our AI Pods catalog.
Some are full builds. Others are small discovery engagements that end with a detailed, costed Work Breakdown Structure you can keep.
You should be able to look at the catalog, get a reasonable sense of the investment, and decide whether talking to us makes sense. That costs us a few discovery calls, and we're fine with it.
Hourly Work Still Has a Place
We haven't declared hourly billing dead. There are situations where it's exactly the right model. If you already have a product team, own the backlog, and simply want experienced engineers working under your direction, our Staff Augmentation model still bills by the hour.
The distinction is straightforward: when we own the delivery commitment, we price the scope in points. When you own the work and direct the team, we bill for the time you're using. Forcing both situations into the same commercial model would be just as artificial as pretending hourly billing was always wrong.
What Comes Next
AI Pods are still evolving. We'll keep expanding the catalog as we see more repeatable project shapes. We'll publish more of the engineering standards behind the model as they mature.
More importantly, we want to publish the numbers that show whether the system actually works: first-pass acceptance, rework, points committed versus delivered, and defects that escape into production. We won't publish those numbers until they're real. Until then, the space stays empty.
The Bigger Change
Koombea has been building software since 2007. AI Pods don't change what clients ultimately hire us to do. They change what we ask them to buy.
Instead of buying our time and hoping the estimate holds, you buy a defined outcome at a price you know before we begin. You get the scope, we take the estimation risk, and how we sell finally matches what clients wanted all along: the finished product, not the hours behind it.
If you're weighing a build and want the number before you commit, book a Priority Sync and we'll scope it with you. If a call still feels premature, the AI readiness assessment is there when you're ready.
