👉 Idea to reality in 5 days! - read our teams developer diary!
Fynlo is a bold startup tackling one of the transport industry’s hardest problems: carbon emissions. Their vision is clear: make it radically easier for fleet operators to track, understand, and act on their CO₂ impact. But unlike traditional solutions that drown users in data or trade in vague offsetting schemes, Fynlo’s proposition is refreshingly concrete. It’s a smart layer that plugs into existing fleet systems, calculates real emissions, and gives managers the tools to reduce or reinvest, all in one interface.
At the heart of the platform is an AI assistant named Fynlo, guiding users with friendly prompts and sharp insights. The team needed a prototype, and fast, to showcase this vision to enterprise clients and investors at an upcoming trade show.
The clock was ticking. Fynlo had secured interest from household names in logtics, retail and transportation. These are prospects with fleets of over 40,000 vehicles. What they didn’t have was a working demo. They needed a clickable prototype that could tell the whole story in under five minutes. No fluff. No filler.
The prototype had to look real, feel smart, and simulate live data, without relying on complex integrations. And the stakes were high. The goal was to land letters of intent from customers and unlock the next funding round. Fynlo needed to impress.
This AI-native approach changed the rules of what was possible in five days
We kicked off with an intensive design sprint: a one-day in-person workshop, followed by a virtual deep-dive and three days of rapid build. We worked shoulder-to-shoulder with Fynlo in the open, ideas-first, no-hierarchy style that defines how we operate. From brand voice to data flow, every decision was collaborative.
Rather than hand-waving around AI, we showed it. The prototype included a working chatbot assistant powered by GPT-4 prompts, surfacing real suggestions from synthetic data. Users could ask things like “How’s my fleet doing this week?” or “Which vehicles are underperforming?” and get instant, helpful responses.
The offsetting flow, a key differentiator, was built with clarity and intent. Instead of abstract carbon credits, fleet managers could “spend” their emissions locally. One click, one certificate, one visible impact. For the prototype, we featured a major health charity, embodying Fynlo’s ethos: invest where your drivers live, not where trees are cheap.
We used LLMs to transcribe and synthesise workshop outputs in real time, helping the team move from scribbles to structured requirements in hours, not days or weeks.
To simulate integration, we created a fake telematics dashboard with a simple “Open in Fynlo” link. This visually demonstrated the ease of adoption. Behind the scenes, we used Next.js, Supabase, and Vercel to power a hosted, polished web app that felt production-ready, even if the data was synthetic.
We focused hard on UX: clean, scannable dashboards, accessible colour schemes, vehicle-level drill-downs, and a consistent visual language. Every interaction was optimised for time-poor fleet managers who don’t want to read manuals. Just see, act, and move on.
AI wasn’t an add-on. It was foundational. From planning through design to delivery, intelligent tooling accelerated the entire process. We used LLMs to transcribe and synthesise workshop outputs in real time, helping the team move from scribbles to structured requirements in hours, not days or weeks. Concepts were validated and iterated with natural language prompts instead of heavyweight specs. Even the chatbot in the prototype was powered by AI-generated data and responses, simulating the end-user experience long before the full backend existed.
This AI-native approach changed the rules of what was possible in five days. Instead of static wireframes or clunky mockups, Fynlo walked away with a fully hosted, clickable experience. With AI supporting everything from emissions calculations to design language, we collapsed traditional build timelines into a rapid, co-creative loop and still kept quality high.
Fynlo wasn’t just built fast. It was built differently. Several key principles of AI-native software played a critical role:
AI participated throughout the lifecycle. It captured transcripts, generated requirement docs, proposed UX flows, calculated CO₂ from synthetic data, and simulated the chatbot. The AI wasn’t just assisting. It was co-developing.
We replaced brittle logic with flexible pipelines. Prompt-based interactions powered the chatbot. Background jobs, seeded by LLM-generated synthetic fleet data, ran estimation calculations. Rather than lock in a rigid system, we scaffolded around adaptability and iteration.
The development process was a dialogue. Developers bounced design sketches off GPT-4, generated sample datasets, and used AI to critique early UI ideas. Designers used Subframe and AI-enhanced tools to produce screens with minimal lag between ideation and implementation.
The sprint surfaced new roles in practice. We had de facto Prompt Engineers crafting chatbot tone and interactions, AI Evaluators tweaking system prompts, and even a Knowledge Engineer managing the carbon calculation logic. These roles weren’t labelled, but they emerged organically.
The result was a prototype that didn’t just look smart. It was smart, and it pointed toward a system that could continue learning, reasoning, and adapting.
In under a week, we delivered a functional prototype that made Fynlo’s promise tangible. Users could:
See their fleet’s emissions at a glance
Drill into individual vehicle performance
Get proactive AI-driven tips from Finlo
Choose to reinvest their emissions into a local cause
Receive a confirmation and certificate instantly
The prototype helped Fynlo crystallise their story. From “carbon tracking” to “carbon intelligence.” It transformed the sales pitch into something real and fundable. Investors saw product-market fit. Enterprise clients saw low-friction adoption. And most importantly, the Fynlo team saw their idea come to life.
The prototype was just the beginning. Now comes the scale-up. Ingesting real telematics data. Refining emissions calculations with third-party accreditation, such as the Smart Freight Centre. And building out Fynlo into a full AI co-pilot, not just a chatbot.
The result was a prototype that didn’t just look smart. It was smart, and it pointed toward a system that could continue learning, reasoning, and adapting.
Fynlo’s platform has the potential to shift how fleet emissions are managed. From a compliance checkbox to a strategic lever. And that’s a challenge worth building for.