The rest of the economy

Most of the economy does not run on software built for it. It runs on freight brokers and property managers, contractors, carriers and distributors: businesses whose most important work happens between the systems they already own. A customer calls. A quote arrives by email. A document needs checking against another. None of these steps belongs to any one tool, so people carry them, holding the context in their heads.

We believe this is where AI will matter most, and where it has so far reached least. The models can now listen, read, write and reason well enough to handle the work between systems. What they lack is an understanding of how a particular business runs: its rules, its exceptions and the judgements its people make every day.

That understanding cannot be gathered from the outside. It is learned by working alongside the people who hold it. So we begin with paid work inside an operating business, against its real workload. The operator pays us to solve a defined problem, which keeps us honest about what is worth building, and the work teaches us the rules and exceptions no specification would capture.

When the same need appears in one business after another, we turn its reusable core into a product for the businesses that come next. Implementations are where our products are found; products are how the work reaches further.

We hold the software to two principles. It should fit into the tools people already use, in their phones and their inboxes, adding a dedicated interface only where the work needs one. And it should be clear about its limits: which actions it takes on its own, which it prepares for a person to review, and which it leaves alone.

Our portfolio follows the shape of the work. Echo is for communication, the calls and messages through which most work arrives. Helm is for operations, the records and workflows that carry it through the day. Magnus is for decisions, reading the documents and data behind them and bringing each exception to the person who should decide. The three are complementary because the work is: it arrives, it is carried through, and somewhere along the way it needs a judgement.

We intend to grow through repeatable software and through long relationships with operators who understand their industries better than we ever will. If we do this well, the rest of the economy will not have to change how it works to benefit from what AI can now do.