Building AI Personas
Interactive personas — or AI personas — are a contested tool in research right now — and for good reason. Built without real human insight at the core, they can be dangerous. Built thoughtfully, they're one of the most powerful tools I've found. Here's my point of view.
Where they fit
Built to flex, interactive personas can sharpen a pitch, accelerate creative ideation and iteration, activate research across an organization, or stand in as an always-on, queryable audience lens.
Scope and data quality set the boundaries of what they can do. Every use case falls somewhere on this map — where it lands determines how the persona should be built and used.
Narrow scope · Rigorous data
Built for one job
Catches problems in creative before they reach the audience it's meant for. Fast, focused, and grounded in evidence.
Broad scope · Rigorous data
Shared infrastructure
A resource the whole org can lean on: cross-functional workshops, enablement, an always-on reference.
Narrow scope · Thin data
Speed and inspiration
Not certainty. Built to sharpen one moment — it doesn't carry weight beyond that.
Broad scope · Thin data
Not a supported play
Unsupported
Philosophy
I treat AI personas as audience models — tools for stress-testing an idea against a specific point of view, never a substitute for the research that grounds them. AI earns its place at discrete tasks: classifying themes across interviews, spotting patterns across sources, acting as a thought partner during synthesis. It never sets the direction.
Only a researcher with real client context can define who an audience is and what actually drives their decisions. If an output ever feels like a discovery rather than a hypothesis to verify, that's the signal to go back to primary research.
Explore an AI persona
Maya is a Director of Treasury and Payments at a global logistics company. I built her for one of my financial services clients, using primary research (both quantitative and qualitative) to capture her needs, motivations, and pain points when choosing a payment provider.
Used alongside similar personas for this client’s other target audiences, Maya became a shared reference point for the creative team. She lets us pressure-test messaging before launch, instead of waiting for post-launch feedback. Try her out below!
Hi, I’m Maya.
Director of Treasury & Payments
Based in London, manages EMEA & APAC flows for a global logistics company.
Contextual ad
Cross-Border Economy
Faster payments reshape global trade flows
Real-time infrastructure is changing how businesses move money, with major implications for treasury teams.
What do you think of the tone in the ad placement?
Guardrails
I follow these rules of engagement every time I set up an interactive persona.
Clear application
The first message states its purpose and its explicit non-uses.
Research grounding
Every answer requires a search of the knowledge base first.
Auditable & explainable
A mandatory sources section whenever an answer draws on research.
Compliant
No PII, proprietary or sensitive data is ever included in the knowledge base.
Reproducible
Every persona is built to the same standard, so the results are consistent and easy to check.
Curious how to systematize AI personas in your organization to accelerate product/design collaboration or creative ideation? The best place to start is the research you already have: take stock of your existing primary data on a single audience, then pick one narrow job for a persona to do. That tells us where you land on the data quality / scope 2x2, and whether a persona can help right now or needs stronger research behind it first.
I’d love to connect! Get in touch at info@phorasystems.io and we can build your systems together.