People close to the problem, machines at scale.
We're an AI-native software house: experienced people using automation and agents where the machine is better — and people where people are irreplaceable.
The service pays the bills while our own products mature. Whoever talks to you during discovery is who will write the code.
Tensoor comes from tensor.
A tensor is the mathematical object that generalizes scalars, vectors and matrices — and it's literally the data structure every modern neural network operates on. It's what flows through every layer in TensorFlow and PyTorch.
The connection to AI isn't a figure of speech: it's the exact technical term. And the tensor's hierarchy also describes how a project grows here.
What we do every day.
Build software, automation and AI systems for companies that need it and don't have a team to do it — using agents where the machine is better, and people where people are irreplaceable.
Where we want to get to.
In three years, half our revenue won't come from billed hours — it'll come from product and recurring revenue built with what we learned from the service side.
It's our thesis turned into a number, so anyone on the team can check it. If in three years the math still comes out 95% billed hours, we've become an ordinary software house with AI inside — and we'd rather find that out in time.
Small, senior and close by.
A few people, all with real production mileage. Agents handle research, volume and repetitive tasks; people handle decisions, engineering, creativity and the relationship.
In practice: you talk directly to whoever builds, you follow the board, and you never hear "it was the AI" as an answer.
What Tensoor is not.
More useful than the vision at this stage — it's what keeps us from drifting.
"Cheap devs from Brazil" is a global price war — and there's always someone cheaper.
A website is an entry point, not the business.
We use agents where they pay for themselves — and we measure the result.
If everything bottlenecks on the founder, it hasn't scaled.
Values as a decision rule.
A value that doesn't tell you what to do in a concrete situation is just an adjective. Here the format is: when X happens, Tensoor does Y.
The machine handles volume, the person answers for the result
When an agent gets something wrong with a client, the person who owns that flow is the one who answers for it. "It was the AI" isn't an answer — not to the client, not internally.
Real numbers beat a nice-sounding opinion
Price, timeline, margin and agent gains are hypotheses until they're measured. We measure before we scale.
Nothing is sold before it exists
When sales promises what the house can't deliver, that's the house's problem. An agent can help write the proposal; a person is who guarantees the delivery.
Contradictions go to the table, not the hallway
When two decisions contradict each other, it gets said out loud and turns into a recorded decision. An invisible contradiction is the one that ends up deciding on its own.
Price isn't the argument
When a client only wants the cheapest option, we turn it down. What we sell is speed, reliability, specialization and senior oversight.
If you can't explain it, you haven't delivered it
If no one here can explain the system they delivered, it isn't ready — even if it's working. That's what separates engineering from luck.