Market sizing, unit economics and a Monte Carlo baseline computed from cited public benchmarks - plus a deep consumer simulation where 48 AI agents with personas, budgets and memory make real purchase decisions. AI strategy on top. Not a slide generator.
Pick the channels you could realistically run. The model weights spend across them.
3,000 simulated futures. Shaded band: 10th-90th percentile. Line: median. Parameters sampled around cited benchmarks. Instant baseline - no AI agents here. For emergent consumer behavior, run the deep simulation below.
A panel of 48 AI consumers with persistent personas, budgets and memory stands in for the people your marketing reaches over 6 months. Every purchase, churn and recommendation is a live model inference - not a formula. Takes about a minute.
What this is: an LLM-agent market simulation (method: Generative Agents, Park et al. 2023; EconAgent, ACL 2024; MarketFish, MIT 2026), calibrated with base-rate priors from cited conversion benchmarks. What it is not: a validated predictor. LLM consumer panels systematically skew optimistic, so treat the curve as directional behavioral evidence, not a forecast.
| Channel | Spend/mo | Cust/mo | CAC | Ramp |
|---|
Scenarios recompute the model instantly. The AI narrative stays tied to the original run.