The Hypothesis

Can structured, machine-queryable customer personas replace the slide decks and PDFs that no system actually uses?

The Concept

Every organisation has customer personas — but they live in slide decks, PDFs, or someone's head. They're disconnected from the systems that produce marketing content. This experiment builds personas as structured, queryable data objects: demographics, motivations, pain points, objections, preferred tone, buying triggers, channel affinities. Exposed through a protocol layer so any AI agent, automation, or business user can connect, query, and interact with them. Not documents — live infrastructure.

Persona and cohort builder

The hypothesis

Can structured, machine-queryable customer personas replace the slide decks and PDFs that no system actually uses?

The concept

Every organisation has customer personas — but they live in slide decks, PDFs, or someone’s head. They’re disconnected from the systems that produce marketing content. This experiment builds personas as structured, queryable data objects: demographics, motivations, pain points, objections, preferred tone, buying triggers, channel affinities. Exposed through a protocol layer so any AI agent, automation, or business user can connect, query, and interact with them. Not documents — live infrastructure.

How it works

1. Define personas — rich attributes: demographics, motivations, pain points, objections, tone, triggers
2. Map real customers — connect CRM/analytics data, map customers to persona segments
3. Expose via protocol — structured API for any agent or tool to query cohort data
4. Feed downstream — landing pages, emails, ads, lifecycle flows all consume persona context
5. Refine from outcomes — behavioural data flows back, personas auto-update

Personas become infrastructure — a living, queryable layer that every tool in the stack can plug into.

What it explores

• Do structured, machine-queryable personas produce better marketing output than document-based ones?
• Can AI agents reliably use persona context to tailor content for specific segments?
• How many attributes does a persona need before it’s useful to an AI agent?
• Can the system discover new segments from behavioural clustering that don’t match existing personas?
• Does conversational access (business users asking questions in natural language) drive adoption?

What we found

• Teams using structured personas saw a 28% improvement in campaign relevance scores
• AI agents querying persona data produced more targeted content than agents given generic instructions
• Behavioural clustering discovered 3 new segments the team hadn’t identified manually
• Conversational access drove 4x more persona queries than the previous dashboard-based approach

Learnings

• Personas are only useful if systems can access them programmatically — the protocol layer is what makes them infrastructure, not documentation.

• Weight pain points and objections over demographics in persona schemas — they shape messaging directly and are what agents actually use.

• Treat personas as living objects with expiry dates — stale personas are worse than none because they create false confidence.

• Conversational access drives adoption — lowering the query barrier matters as much as the quality of the data behind it.

Where it goes next

This experiment and the MCP protocol experiment are converging. The structured persona layer is becoming a core component of DataSpec, with the MCP protocol providing the query interface. The open question: how do you version personas over time so you can track how your understanding of a segment evolved?