Introducing Recommend Rosetta
Recommend Rosetta is a family of layers on top of large language models. They narrow the gap between what a general model knows and what your business needs, and deliver work your team would sign off.
Know-how is behind every call
Take a retailer with stores along the coast, and one question: how much should we order before a heatwave? The forecast is only the start. The answer depends on which weekend the heat lands on, how many tourists are already in town, what sold out last summer and what ended up as waste, and how quickly a delivery can actually reach each store.
Or a distributor whose competitor has just cut prices by 10%. Which customers do we defend? It depends on who is really price-sensitive and who stays for the service, which accounts carry the margin, and whether the cut is a short promotion or a new price level.
Or a pharma company preparing a launch. Which pharmacies come first? Not simply the biggest ones, but those where patients, prescribers and location make the first weeks count.
None of these answers sits in a single report. Each depends on context that is hidden, changing, and spread across systems and people.
Where general models fall short
The people who answer questions like these well have know-how. They use it every day, often without spelling it out, and it is a big part of why they are good at what they do. That is also why a general model struggles to keep up with them. A prompt only goes so far, and no model starts reasoning the way they do on its own.
General models are remarkable at reasoning and language, but inside a business they keep missing the same things. They don't read how people and markets actually behave. They don't make the call the way an expert in the field would. They don't check what they are given as carefully as they should. They don't write in your tone and style. And they often misread your context: one small detail taken the wrong way, and the whole output goes off course, after a lot of time and money have already gone into it.
Introducing Recommend Rosetta
Today we are introducing Recommend Rosetta, a family of layers: layers of intelligence you own that start where the general model stops and shift AI outputs closer to the final outcome.
Each layer is built for one part of that gap: how people and markets behave, how a call gets made, what needs checking, how it should sound, and what context matters. They work together on top of the model and get better as your teams work with them, so the gap gets smaller with every task, and your experts spend their time on the calls that need them.
What Recommend Rosetta is
A general model is like a very capable new hire. It knows a lot, writes well and can reason through almost anything. What it doesn't know yet is how your business works: how your team makes a call, what they check, and what they would never sign off on.
Recommend Rosetta closes that gap. Planning stock, responding to a competitor's price and sizing an offer are different kinds of work, and each needs its own expertise. Each gets a layer of its own, so instead of a generalist's answer you get work done the way your best people would do it.
Technically, a Rosetta layer is a specialised component on top of a general-purpose foundation. Each layer is scoped to one kind of process, and a task goes to the layer built for that kind of work. The foundation brings the general capability and the layer brings the specialisation, so the output reflects how that kind of work is actually done in your business.
One family, many layers
What it means for you
Ready to use. Output your experts can sign off on, not a starting point.
Your experts stay in charge. Every call they make, and how it turns out, makes the next output better.
It is yours. The layer is shaped by your know-how and your outcomes: intelligence you own.
Works with the LLM you choose. Inside our agents, or connected to your own tools through MCP or an API.
Runs where you choose. On our servers or on your own infrastructure.
