GUST exists because the team behind it has lived through what happens when smart software meets real organisations. Drift, rework, audit gaps, and quiet failure. We are building the system we wished existed when we were the ones on the hook for it.
GUST 1.0 was a multi-agent system for refactoring legacy code. Paying enterprise customers used it on real production workflows. What we learned from those deployments is the reason GUST today looks nothing like what we set out to build.
01 · GUST 1.0
GUST 1.0 was a multi-agent system for refactoring legacy code, deployed with paying enterprise customers, running real production workflows. Not a demo. Not a benchmark. Customer code.
A SaaS company in Deloitte's Finnish Fast 50 used it to convert hundreds of React Native files into React Web for their US client in half the time. A sentiment analytics platform used it to build and deploy a full product in three months. The product shipped. The lesson it surfaced was the more interesting part.
02 · What surfaced
Memory loss was. Lost scope was. Untraced decisions were. Quiet drift was. The same reliability failures behind every stalled enterprise AI project showed up in our own deployments, in front of paying customers.
We had built a refactoring tool. What we kept running into was the layer underneath it. The same layer that would be missing under any agentic AI workflow anyone might want to deploy. The interesting question was never the code. It was whether agentic AI could be governed end to end, in production, against the policies of a real organisation.
03 · GUST today
GUST today is the product those deployments told us to build. Not another agent. Not another framework. The coordination layer underneath them. Scopes. Persistent context. Propagation rules. Audit by default. The substrate that lets any agent, in any framework, be governed end to end.
That is what the rest of this site is about. This page is about the people who got us here.
A CEO who has been on both sides of enterprise AI. A CTO who built the laboratory the product came out of. A CRO who has scaled deep enterprise software before.
Michael has run venture-backed and private-equity-backed scale-ups across the US and Europe to three successful exits: Telarix (acquired by TOMIA, now a standard in inter-carrier telecom settlement), Sentori, and IPsoft.
The IPsoft chapter matters here. It was one of the earliest enterprise AI and automation companies, years before agentic AI was a category. Michael has seen what makes AI work inside large organisations, and he has seen what makes it fail. GUST is built on the second list.
A decade of full-stack engineering across banking, consumer apps, and mixed-technology delivery. Jurmen leads engineering at GUST and built the multi-agent code refactoring system that became GUST 1.0.
The decisions he made there about memory, scope, propagation, and audit are the foundations the current platform is built on. He is the person responsible for the difference between an agent that demos well and an agent that survives a production deployment.
Twenty-five years inside multinational software companies, including Adobe and MobileBridge, with revenue responsibility measured in tens of millions. Mika has sat on the boards of four high-growth scale-ups.
He owns commercial strategy at GUST: building the partnership and channel infrastructure that turns deep enterprise software from a few good pilots into a real business.
These are not slogans. They are the lessons GUST 1.0 left us with, and the reason GUST 2.0 looks the way it does.
Capability gets the press. Reliability gets the production deployment. The frontier models are remarkable. The reason agentic AI is not trusted in high-stakes workflows is not that the models cannot reason. It is that nothing remembers, nothing coordinates, and nothing can be audited. We work on the second list.
Procurement, compliance, and legal will block any AI system that cannot show its work. We treat governance as a property of the system, written into the type model. Every scope is auditable. Every change is logged. Every policy is enforced structurally, not by hope.
AI vendors talk about their models. We think the more useful model is your business: how teams depend on each other, which decisions cascade, which policies apply where. That model is what GUST keeps current. Every agent you run coordinates against the same evolving picture of how your enterprise actually works.
Four people who have run the buying side and the building side of enterprise technology. They keep us honest about what enterprise customers will actually deploy.
GUST is opening access to selected alpha partners. We are looking for teams who want to make agentic AI reliable, governed, and useful in real workflows. We are also speaking with investors building conviction in this space.
