Most founder stories start with a bolt of inspiration in a coffee shop. Mine started with a stuck delivery at 2am GMT, a shift lead in Leipzig, and twenty years of watching enterprise software fail the people who actually use it.
The long prologue
I spent two decades inside SAP EWM, TM, and S/4HANA. I wrote a book on it. I ran go-lives on three continents. I saw warehouses of every size, every industry, every quirk.
There is a pattern you notice after the hundredth implementation.
Enterprise software is very good at storing the state of the world. It is very bad at reasoning about it. The reasoning always ends up in someone's head — a senior consultant, a shift lead, a compliance officer who has been around long enough to know where the bodies are buried.
Those people are expensive, rare, and getting rarer. Younger operators do not want to spend a decade learning SAP transaction codes. Why would they? The knowledge is not written down anywhere useful. It is carried.
Why LLMs changed the calculation
The thing that shifted in 2023 was not that AI got smarter. It was that AI got fluent in enterprise context. You could finally hand an LLM a pile of SAP config, a real-world scenario, and a few tools — and the model could reason about them in the vocabulary of the floor.
The work became less about training models and more about building the surroundings: the memory, the tools, the governance. That is software engineering. I know how to do that.
More importantly, the problems I had spent twenty years watching suddenly had a shape that was solvable.
A stuck delivery is a pattern-matching problem against a thousand previous stuck deliveries. An EPR reconciliation is a cross-reference problem across three systems. A wave failure is a diagnostic tree with well-defined branches. All of these are exactly the shape of problem modern agents can handle — if you give them the right memory and the right tools.
The pivot
Futuryntix is a rebrand, not a new company. The old Futuryntix did SAP consulting and published a book on EWM. The new Futuryntix builds AI agents for enterprise supply chains.
The logo changed. The registered company (UK, Companies House) did not. The twenty years of domain knowledge did not. The opinions about how warehouses actually work did not.
What changed is that the product is software, not days.
Why the traditional AI crowd will be late
The people best placed to build this are not pure AI researchers. They are enterprise software people who understand both what LLMs can do and what warehouses, factories, and procurement departments actually need.
That is a narrow Venn. Most AI companies lack the enterprise fluency. Most enterprise companies lack the AI fluency. The ones who sit in the middle have a quiet, compounding advantage.
That middle is where Futuryntix was built to sit.
What this means for anyone else
If you are a senior practitioner in a deep enterprise domain — SAP, Oracle, Salesforce, industry-specific systems — and you have been wondering whether the AI wave is for you: it is.
You do not need to go back and learn transformer internals. You need to learn how to build agents, evaluate them, and ship them. That is a shorter path than the one most researchers are on in the opposite direction.
The AI-first era of enterprise software will not be won by people who discovered the enterprise last year. It will be won by people who know both sides — and there are fewer of us than you might think.