Selling a result, not a tool

Over the years at Startup Bakery, we have signed quite a few contracts and watched the same scene play out more often than I’d like to admit. The demo goes well, they like the product, the deal is signed. Then three or four months go by, you check the usage data, and you discover that the client has practically never logged into the platform. The contract is there, and so is the revenue. The problem you were supposed to solve for them, however, is exactly where you left it.

You had sold them a tool. And they didn't really care about the tool. They wanted the problem to go away, and using the tool to make it go away was a job that still fell on their shoulders, requiring time and skills they often lacked.

The tool was never the product

Software has always sold the same promise: I give you the ability to do something better and faster. The catch is that the ability isn't the result. Between "I give you the system to keep expenses in check" and "your expenses are in check," there is a whole lot of heavy lifting the client still has to do. And let's be honest, most of the time, they just don't do it.

For years, this was a technological limitation, not a choice. We couldn't deliver the result, so we delivered the closest possible tool and crossed our fingers. With AI agents, that barrier is crumbling: the work you used to offload onto the client can now be kept in-house. This is the direction we are heading at Startup Bakery, and it is much more than just a change in slogan.

What changes in pricing

When you sell a tool, the price is anchored to access. A certain amount per user, a certain amount per month, maybe three tiers, and that's it. It's convenient, it's predictable, and it's the reason we've gotten used to thinking in terms of licenses.

When you sell a result, that anchor drops. The client stops buying "five seats" and starts paying you because you manage the invoices instead of them. What they are buying is the outcome; they couldn't care less about the software itself.

And usually, you can charge more for the outcome. Not because you're expensive, but because you are taking on a chunk of work that the client previously handled internally, and that work has a cost they know perfectly well: it's hours from someone on their team. Negotiating that number is much easier than explaining why your subscription costs 49 euros instead of 39 euros.

There are many ways to package it: a price per single result, a cut tied to the savings you generate, a fixed fee for a guaranteed outcome, or a mix of these. Which one works depends on how measurable that outcome is and how clearly it can be attributed to you. And this is where things get slippery.

You take the risk in-house

Truly selling the outcome means bringing something inside that, with the tool, stayed outside the door: the risk.

With a tool, the execution risk remains with the client: if they use it poorly, that's their problem. With a result, that risk becomes yours. If the agent messes up, if the output needs to be double-checked, if the outcome doesn't arrive within the promised timeframe, you can no longer say that they didn't configure it right. You simply didn't deliver.

Two practical consequences follow from this. The first is that the price must bake in a margin for the risk you are shouldering. The second is that there has to be a machine behind the scenes backing up the promise: output quality controls, monitoring, and someone or something that notices when things derail. Without that machine, the outcome-based model is just a refined way to lose money.

Putting it in black and white

A rule I learned the hard way is that the outcome must be written in a measurable way, otherwise the contract is just a postponed lawsuit. "Improving efficiency" means nothing. "Invoices categorized and reconciled without manual intervention, and renewals flagged before they trigger" means something.

A few things need to be nailed down beforehand, not argued over after the damage is done. You must define what counts as delivered, where the baseline for measurement starts, and what happens if the outcome isn't reached. And you need a clear boundary of liability: if the client hands you the wrong data, the wrong result cannot fall on you. It seems obvious, but it's exactly the kind of thing you'll be tearing each other apart over six months later if you didn't set it in stone from the start.

Then there is something trivial but decisive: SLAs go on the result, not on uptime. The client doesn't really care if the platform is up 99.9% of the time. They care that the result arrives.

A concrete case: Kontai

An example we have in-house is Kontai, one of the startups from our studio. Kontai manages a company's recurring expenses: software subscriptions, utilities, licenses, and all those costs that multiply and that almost no one manages to keep together, because they are scattered across emails, invoices, company cards, and contracts people forget they even signed.

We all know the "tool" version of a product like this: a dashboard where you upload expenses, categorize them, and keep an eye on renewals. Tidy, nice to look at, and sharing the same fate as so many dashboards: you open it the first month and then never again. Because keeping it updated remains a chore on the shoulders of the finance team, who simply don't have the time.

Kontai is taking the opposite path. The AI captures invoices from the email inbox and payment systems, reconciles and categorizes them on its own, sends alerts before a renewal hits, and flags anomalous charges before they go unnoticed. The client doesn't have to "use a tool": the product does the work. What they are left holding is the result, meaning actually knowing how much they are spending and no longer discovering renewals when they have already triggered.

You can also tell from how it's pitched to the client. You don't start with "try the software," you start with a free assessment that captures how mature or immature the company is in managing its recurring costs, with concrete steps to take. First the outcome, then maybe the features. And that is where, over time, the price can latch onto what the client really cares about: the spending they manage to keep in check and the money they put back in their pockets.

In conclusion

Ultimately, the product was never the software. It has always been the result. It's just that until yesterday we didn't know how to deliver it, so we sold the tool and hoped the client would do the rest. Today, piece by piece, we can start doing that "rest" ourselves.

What changes, deep down, is what's in front of you when you sign: no longer the promise that the client, if they try hard enough, will manage on their own, but the already solved problem. It is a more uncomfortable job than the old one, it takes on more risks, and offers fewer shortcuts. It is also much harder to copy, and that is exactly why it makes sense to get into it now, while almost no one is doing it for real.