AI-Native, by Design

From AI first to AI native

Two years ago we wrote that we would take an AI-first approach. It was the right intention for its time: look at every problem and ask where AI could help. Since then the question has changed. AI is no longer a feature to be added to a product. It is a way the product works.

That is what we mean by AI native. In our products, AI is in the architecture: in how a visitor is recognised, how a question is answered, how a pattern is noticed, and increasingly in how a task gets done.

What we have already shipped

• A conversational assistant with a face, deployed at information points for a Dubai government utility. It recognises whether the person in front of it is an employee or a visitor, greets them in their own language, answers questions and changes what is on the screen by voice.
• Audience-aware content for a Dubai healthcare provider, where the screen responds to the person looking at it.
• Face registration and recognition at the kiosk in BonVoy VMS, with consent, so a returning visitor is known at every touchpoint.
• Lobbi, an Arabic and English AI receptionist that checks visitors in, tells the host and shows the way.
• Analytics and prediction on the data that visitor and elevator journeys generate.

Why now, and why here

• The region has decided. The UAE's national AI strategy sets the ambition, and government and enterprise clients now ask how a system uses AI before they ask what it costs.
• Sovereignty matters. Our clients need AI that runs in region, on infrastructure they control, on data that is never used to train someone else's model.
• Arabic is not an afterthought. An assistant that serves this region must listen and speak in Gulf Arabic and work right to left.
• AI has learned to act. Open standards such as the Model Context Protocol let assistants use systems, and that makes governance the main engineering problem. It is why we built Limigate.
• Buildings must do more with less. Fewer people at the desk, less energy on the wall, more accountability for every square metre.

Our guardrails

• Grounded or silent: an assistant answers from approved knowledge, and offers a person when it has none.
• Human in the loop: actions are confirmed by the person they are for, and safety and access decisions stay under human supervision.
• Permissioned: AI acts with the rights of its user, and every action is logged.
• Explainable: decisions can be traced to their source.
• Sovereign: data residency and in-region hosting by default.
• Private by design: no more data than the task needs, kept no longer than policy allows.
• Never a dead end: if the AI is unavailable, the building still works.

Our method

• Discover: find the decision or the conversation where intelligence changes the outcome.
• Pilot: prove it on one site, with real users and real data.
• Productise: build what worked into the product, so every client benefits.
• Operate: measure it, support it and keep improving it under a service level.

AI in how we work

Our engineers use AI assistance every day, in design, coding, testing and support, with human review of everything that ships. It is one reason a small UAE team can maintain a portfolio this broad. Read more about how we build.