What we aren't
Another company bolting a chatbot onto spreadsheets and calling it AI.
We use large language models for ingestion and narration, turning a simulation into a written brief. The number underneath comes from a structural model with its own balance sheets and behavioral equations, not a language model's best guess.
A dashboard with more charts than the last one.
A dashboard shows what already happened. This lets you change an assumption and trace the consequences through 115 industries, or 51 states, in the time it takes to read the result.
A consultancy that delivers a report and moves on to the next client.
Consulting that doesn't expire. The model we build for an engagement doesn't get archived when the invoice clears. Clients keep querying it, long after the report is delivered.
A black box that spits out numbers you're supposed to trust.
Every mechanism is certified against a registered prediction before it is switched on, and every solved year passes accounting checks that hold stored positions to within one part in a million. Baselines are built on the CBO, Treasury and OECD outlooks. When our China model's current account surplus overshot the World Bank's projection, we rebuilt the baseline until it matched. The calibration constants and their sources are published on the model page.
Phylleos