Open Food Chain builds product passports for the food industry: scan a code on a bar of chocolate or a bottle of olive oil and see where it came from, batch by batch, with the evidence behind every claim. Evidence, not a claim. And that evidence chain did not start as a product. It started as an idea, with no architecture, no security review, and no steady way to run the work beyond a handful of people talking.
This case is about the distance between those two states, and how it was closed: first a foundation, then an audit, then agentic development on the Runbaze platform. One idea, three engagements, one live product, with a record their biggest clients can inspect.
Their platform turns supplier certificates, batch records, and origin data into one passport. A shopper can scan it. So can a buyer. So can a regulator. And the categories where this matters most are the ones where origin decides the sale: cacao, olive oil, fish, tea, juice. Their users are food brands whose buyers include some of the biggest names in the industry.
None of this is trivial engineering. A passport platform sits at the intersection of messy inputs and unforgiving outputs. Evidence arrives from many suppliers in as many formats: certificates, audit reports, origin declarations, geolocation data. It has to be verified, structured, and bound to the right product at the right batch. The output end is consumer-scale: a QR code on a supermarket shelf, resolving on the GS1 Digital Link standard for whichever batch is in the shopper's hand, with analytics behind every scan and enterprise clients who want it connected to their own systems. The roadmap that serves all of that is never short.
Two dates turned that roadmap urgent. From 27 September 2026, the EU's Empowering Consumers Directive bans generic environmental claims: "natural", "eco", "climate neutral" are no longer allowed on a label without recognised proof behind them. And from 30 December 2026, the EU Deforestation Regulation requires large companies placing cocoa, coffee, soy and other listed commodities on the European market to prove, with geolocation data down to the plot of land, that their supply chains are deforestation-free.
For Open Food Chain's clients, product evidence is turning from marketing into law. For Open Food Chain, that meant every feature on the roadmap was wanted at once, and there was no room left for a slow start.
Impact first is a rational bet for a young company. Chase the market, get something in front of customers, worry about the plumbing later. Fast iteration and fast demos, that is exactly what this approach is built for.
Then the product grows and the bill arrives. Technical debt compounds, scaling gets expensive, gaps in security surface, and nobody can point to one source of truth. What used to be speed becomes drag. And it stops being a product problem the moment it becomes a company problem.
None of that is a talent story. The concept behind the Product Passport was strong from day one, and the team understood food supply chains, and the buyers behind them, better than most engineers ever will. What they did not have yet was a system. The people who understood the product best were not always the people who could carry it safely from idea to production. Structure is the part you can install, and that is where the work with Runbaze began.
Two states, in order: what the idea looked like on its own, and what it looked like once a foundation, an audit, and Pulse sat underneath it.
Most companies do not lack talent. They lack a system. Work happens and features get built, until something breaks and nobody can say why, or who owns the fix. That is a company without a project management framework: busy, capable, and quietly at risk. So the first engagement touched no features at all, on purpose.
Without structure, work happens in bursts. Something urgent lands, everything else stops, and priorities get decided in the moment. A real cadence, with real checkpoints and a shared queue, replaces that with rhythm. Progress becomes visible before the deadline, not after it is missed.
A company without this framework can still get by for a while. But every release carries more risk than it needs to, and every delay traces back to the same root cause: nothing was written down, nothing was systemised. A foundation like this does not slow a company down. It makes speed safe to sustain.
With the foundation in place, the next step was to test it: a full technical audit of how the platform was built, hosted, and run. What followed was not one big fix. It was a sequence of deliberate moves that quietly changed how reliable, safe, and scalable the product is.
The most consequential finding was data access: gaps in how information was separated between the companies using the platform, gaps that could have let one organisation see or affect another's data. That was fixed with urgency. Access rules were rebuilt so each company's data stays inside its own boundary.
A handful of records had also been marked verified without the evidence behind them. A small gap, but a meaningful one for a product whose value rests on traceability. Each case was resolved on its own, then closed for good with checks that make the gap impossible to repeat.
Once the foundation held, the idea could become software without losing the discipline underneath it. The team had adopted AI-assisted development early, and the speed was real. They also saw its ceiling: a chat window helps one developer, once, and leaves nothing behind that a reviewer, let alone an enterprise client, can inspect. So they took development beyond the chat interface, onto Pulse (Human+AI dev), the layer their own development now runs on, deployed inside their environment and working against the systems they already run.
Their development now runs the way their supply chains do. Work goes in as a specified brief, not a prompt in a tab. Agents execute against it, chained by autopilot, inside safety rules and cost budgets. A person validates the decisions that matter, at gates where a wrong call would cost something. And every run, every approval, and every change lands on the record, from intent to deploy. Four stages, always in this order: specify, execute, validate, record. Fast, governed, and on the record.
What this buys is parallelism. A specified brief can be worked by an agent while the team holds judgment: more of the roadmap moves at once than they could ever staff, and nothing ships without a gate a person owned. When a large client, or a regulator behind that client, asks how a feature came to be, the answer is not a shrug toward an AI tool. It is a record: this brief, these runs, this review, this deploy.
What came out the other side is Product Passport, a live system, not a draft. Against the original idea, it is more flexible: built to work across categories, from cacao to olive oil to fish, not one product type. It is more interactive: a shopper, a buyer, and a regulator each get their own view of the same evidence. And it carries more complete information and more complete data than the idea ever could, because the foundation underneath was built to hold that much.
There is a symmetry worth naming here. A company that sells proof of origin should be able to prove its own process too: what was reviewed, what was decided, why. That is what the audit and the foundation gave Open Food Chain, before Pulse ever touched a brief. And it is what the record now extends to every piece of software they ship.
This is what agentic development looks like when it is built for the enterprise rather than the demo: agents on the volume, people at the gates, and a record that answers. If your roadmap is bigger than your team, the sequence starts small and proves itself first. One call, and we will tell you what is missing in your foundation, and whether Runbaze is the right partner for it.