DeBe – A self-initiated, AI-enabled venture
In Brief
Self-initiated venture, no client brief.
AI-enhanced design, AI-enabled journeys.
Led end-to-end: research to go-to-market planning.
Synthetic patient modelling for quantitative depth.
Validated competitive gap across clinical literature.
Six diabetes signals, no platform synthesises them all.
Concept: unified AI recommendation layer across all six.
Four-phase go-to-market roadmap designed and documented.
Outcomes and Impact
537M adults with diabetes globally.
6 health signals, no platform synthesises them all.
4 phases from research to public release.
0 deployed systems with a unified AI recommendation layer across all six signals.
Situation
The tools to manage diabetes exist. CGMs, fitness trackers, nutrition apps, blood pressure monitors. But they were built in isolation. They do not talk to each other. They do not think.
The entire responsibility of interpretation sits with the patient. Every correlation, every behaviour adjustment, every clinical decision is manual. For most people, that is not sustainable.
No existing platform integrates all six diabetes-relevant health signals simultaneously into a unified AI recommendation layer. The closest competitor integrates five. Blood oxygen is absent from every patient-facing product currently deployed. The barrier is structural: fragmented data standards, closed device ecosystems, no shared intelligence across sources.
This is a self-initiated project.
No client commissioned it.
The brief came from the problem itself.
Approach
AI compressed weeks of desk research into days. That changed what was possible at the start, when the strategic choices were still open and the problem was still forming.
The competitive landscape was validated across clinical literature, market reports, and platform data before a single design decision was made. Synthetic patient modelling added quantitative depth, simulating population-scale behaviour patterns and clinical edge cases where real recruitment would be too slow. Qualitative depth interviews with long-term patients provided the human signal no model can replace.
Two behavioural segments emerged. The Finders are newly diagnosed patients, still forming habits, looking for structure. The Fixers are long-term patients who know exactly where the current tools fail. Launch focus is the Fixers: faster validation, denser signal, clearer evidence base.
Regulatory obligations across GDPR, AI Act, EAA, and DSA/DMA were mapped from the start and embedded into the design system.
AI ran through every stage of this work, but at two distinct levels. As a strategic capability, it accelerated the thinking before any design decisions were made. As a product decision, it became the operating logic of the service itself: the mechanism that connects six signals, surfaces patterns, and generates actions the patient can trust.
Keeping those two roles distinct was one of the more consequential design choices in this project.
System
DeBe operates across three layers.
Sense is the device-agnostic data layer, integrating across CGMs, smartwatches, blood pressure monitors, oximeters, and fitness trackers via Apple HealthKit, Google Health Connect, and FHIR. No lock-in.
Understand is the experience layer. A mobile app for real-time awareness. A web interface for history, trends, and clinical sharing in plain language.
Act is the AI agent layer. Meal suggestions calibrated to live glucose levels. Movement prompts timed to trend data. Shopping list adjustments. Pre-formatted GP reports. Alerts before a hypo develops, not after symptoms appear.
The long-term ambition is to solve the interoperability problem at its root: open standards, shared data infrastructure, a system that works across every device a patient already owns.
AI is not a feature added to the system. It is the operating logic connecting all three layers.
Outcomes
This project is in progress. The outcomes at this stage are methodological proof points, not commercial results.
Competitive gap confirmed and evidenced across peer-reviewed clinical literature, NIH/NIDDK research, and OECD health technology analysis.
Six-signal integration identified as the primary structural market gap: SpO₂ is absent from all existing patient-facing platforms.
Two behavioural segments defined with distinct onboarding strategies and one shared product architecture.
Regulatory compliance framework mapped across GDPR, AI Act, EAA, and DSA/DMA.
Four-phase go-to-market roadmap: Understand and Define → Prototype and Validate → Ship and Learn → Scale and Evolve.
Ale