Adapta model routing strategy
Max Peters, CEO of Adapta, on building AI agents for Brazilian SMBs
The real moat is not a single proprietary model, it is the routing system that decides when to spend more and when to spend less. For Adapta, that matters because SMB customers pay fixed subscriptions but use AI heavily, so margin depends on sending each task to the cheapest model that still gives a good enough answer. That turns model selection into a core product and economics layer, not just backend plumbing.
-
This is the same move other multi model apps have made. Copy.ai described switching between GPT 4 and Anthropic models as prices and quality changed, then building one layer above the labs so it could swap in a better model fast. Adapta is doing the same for Brazilian SMB workflows.
-
The routing job is especially important because Adapta is selling one workspace for chat, internal tools, agents, and automations. A simple writing task, a legal document review, and a background workflow do not need the same model. Picking one default model for everything would either hurt output quality or crush gross margin.
-
A close comparable is Langdock in Europe. It routes requests across 40 plus models and charges both seats and usage markup. Adapta is aimed at SMBs rather than regulated enterprises, but the underlying pattern is similar, the workspace that controls model choice can capture value even when the frontier models keep changing.
As models keep getting cheaper and more specialized, the winners in AI workspaces will look less like single model apps and more like traffic controllers for work. That favors Adapta if it keeps turning Brazilian business tasks into test data, because every routed prompt can improve its pricing, defaults, and eventually its own smaller task specific models.