Multi-agent strategy engine
Stratus
A strategy war-game where AI agents fight out a decision before you make it.
Stratus lets you play out a competitive decision instead of just arguing about it. You pick a role, the rest of the company and your competitors each run as their own agent, and they argue toward a move you could take into a real meeting. The public demo is live and runs on synthetic data.
How a round plays
Pick who you are playing: a local brand GM, a group MD, a regional brand lead, or the regional CEO. Each one wants something different out of the same budget.
Everyone bids for a share of one budget while the competitor agents respond. You get back the move that wins the argument, what each of the others pushed for, and the evidence that would change their mind.
Competitors by job, not category
Competitors are grouped by the job a shopper is trying to get done, not by product category.
So it flags a threat from the next category over, the kind a category-by-category view tends to miss.
Honest about its limits
It keeps a ledger of its own past predictions and scores them out of sample. When it gets one wrong, it adjusts.
The numbers come from a contributed industry panel, so everything is share of panel, not real market share, and the app says exactly that on screen. The public demo only ever uses synthetic data. No real brand data goes near the repo.
Stack
The public demo is synthetic-data only; real panel data never goes near it. The part I actually care about is the strategy engine underneath, and this lets you poke at it.