Client
High-volume messaging team
Service
Implement
Status
Shipped
Stack
Persona generation · sequential testing · two-model design
A campaign team's synthetic focus group.
AI-generated personas compare message variants and explain their responses. The simulated results support early review before testing with real customers.
The brief
A team sending high-volume customer messages wanted a way to compare drafts before testing them with customers. The aim was to identify potential reactions and differences between variants earlier in the writing process.
What we built
We built a tool that generates AI personas from audience characteristics supplied by the team. Each persona produces a simulated reaction and vote on the message variants, with results displayed as the run progresses. These responses are generated by models; they are not observations of real customers.
Variants2–4 messages
PersonasAI-generated
React & voteper persona
Testearly-stop
Comparison + reasons
Two models · one generates reactions and votes · another synthesizes the results
Testing and interpretation
The run uses a sequential stopping rule to limit additional simulated votes when one variant leads by the configured threshold. Results can be broken down by persona segment, and a synthesis summarizes the reasons generated by the personas. The threshold describes the simulation; it does not establish how real customers will respond.
Delivery
The shipped tool lets the team compare variants across hundreds of AI personas and review the generated reasoning before a real send. It supports early screening and discussion, with customer response still requiring real-world testing.
SimulationHundredsof AI-generated personas
Variants2–4tested at once, with early-stop
Design2 modelsreactions and synthesis
