OngoingCase study · AI orchestrationAI that controls outcomes, not settings.
Sencyble makes IoT and AI technology for air conditioners. Since May 2026 we’ve been its AI partner. We design, build and run the AI orchestration layer that turns “It’s sticky in here” into the right result in the room, confirms it worked, and speaks up before something goes wrong.
- Client
- Sencyble, IoT and AI for air conditioners
- Engagement
- Ongoing, since May 2026
- Our role
- AI partner: we design, build and run the AI layer
- Focus
- AI orchestration, proactive alerts, self-service
Why it’s hardAI in the physical world needs more than a chatbot.
People describe feelings, not settings.
The same few words have to become the right result in the room, not just a lower number.
Every machine is different.
Each model supports different controls. The AI may only use what the unit in front of it can really do.
Mistakes can’t reach hardware.
A language model can be confidently wrong. Nothing it suggests runs on a machine unchecked.
One request, end to endEvery request runs the same loop.
Understanding people is the AI’s job. Facts, limits and the machine stay with software. We designed and built both, and run the loop that joins them.
- 01 Hear it Typed or spoken, in plain words.
- 02AI Understand “It’s sticky” becomes a goal: drier, cooler air.
- 03⌗Logic Check the room What the room and this unit are doing right now.
- 04AI Decide The right action for this exact unit, or the right kind of help.
- 05⌗Logic Act and confirm Does it, then checks the room really changed. If not, tries again or says so.
- 06AI Remember Keeps a preference only when it repeats, and people can remove it.
GuardrailsLimits that hold, whatever the AI suggests.
The AI suggests; software decides. Nothing reaches a machine until it has been checked against what that unit can safely do, inside limits the operators control.
- Hundredsof deliberately tricky test situations, from confusing requests to wrong readings.
- Release checksfor truthful answers, physical safety and careful learning.
- Privacy controlsso people can see, export or delete what is kept about them.
Proactive and self-serviceIt speaks up before you ask.
The assistant watches for trouble, explains it in plain words and offers a fix people can run themselves. When a fix isn’t enough, a person takes over.
- Spots a problem earlyNotices when the AC isn’t performing as it should.
- Tells you in plain wordsAn alert on the phone before anyone feels it in the room.
- Offers a fix you can runLike the AC’s self-clean, started from the chat.
- Brings in a personIf that doesn’t fix it, a service request goes to a technician.
- Clean my AC
- Why is it noisy?
- Check my filter
- It won’t connect
- Book a service visit
What we work onThe whole AI layer, not just the model.
The engagementOne team, working as part of theirs.
An ongoing partnership since May 2026: priorities agreed together, and the AI moves as Sencyble’s product does. Our people on it:
AI engineersSoftware engineersDevOpsProductGood fit if you’re
- Building a product with AI at its core
- Connecting AI to devices or real-world systems
- Wanting a senior AI team alongside yours