How to Create & Use Synthetic Personas and Digital Twins

Learn how to design, validate, and deploy synthetic personas and digital‑footprint‑based digital twins so you can get reliable feedback on products, designs, concepts and more in minutes, not months.

Who this is for

Marketing researchers, insights pros, CX designers and managers; UX/UI designers; customer success leaders; product developers; architects; industrial designers; marketing consultants, and agency teams who need credible, fast directional input from hard‑to‑reach audiences.

What you’ll learn

• Sourcing & curation: Build a defensible source pack (public talks, articles, interviews, posts, first‑party transcripts) and document provenance.
• Twin scaffolding: Translate sources into values, heuristics, preferences, and likely trade‑offs; separate fact from inference.
• Persona vs. twin: Composite ICP personas for breadth; individual twins for depth — when to use each and how to combine them.
• Grounded conversations: Run NotebookLM chats that stay tied to sources with citations, not vibes. Prompt patterns that curb hallucinations.
• Validation loop: A/B the twin against real interview transcripts; score agreement, tone, and reasoning; set “go/no‑go” thresholds.
• Ethics & risk: Disclaimers, consent pathways, and IP rules. Where this is appropriate — and where it isn’t.
• Deployment: Drop twins into concept tests, UX walkthroughs, copy checks, messaging grids, and stakeholder briefings.
• Governance: Versioning, change logs, and sunset rules so twins don’t drift as the person or market evolves.

Top use cases

Decision panels

• Create a “dream board of directors” to stress‑test strategy from multiple expert viewpoints.
• Assemble cross‑functional panels (e.g., directors and VP's across multiple silos or specific department leaders across multiple target companies) to reconcile trade‑offs.

Hard‑to‑reach interviews

• Simulate in‑depth interviews with plumbers, surgeons, CTOs, or CEOs when calendars are impossible.
• Pressure‑test questions and guides before you field with humans.

Influencer & account insight

• Pre‑read how a specific executive, analyst, or thought leader is likely to react to a pitch or POV.
• Map hot buttons by function (Finance cares about X, Ops about Y) before a high‑stakes meeting.

Concept & message testing

• Rapidly iterate product names, claims, and RTBs; highlight wording that triggers objections.
• Compare reactions across segments/buying stages without re‑recruiting.

UX/CX reviews

• Walk through flows with representative twins; flag friction, clarity issues, and trust breakers.
• Translate feedback into prioritized fixes tied to segment needs.

Synthetic focus groups

• Convene 4–8 twins for moderated discussions; capture convergences/divergences with citations.
• Use as a rehearsal before running the live group.

What you’ll build in class

• Two working twins (one public figure, one customer archetype) in NotebookLM with a documented source map and disclosure page.
• One composite ICP persona set (3–5 segments) with buying‑stage overlays.
• A validated prompt kit: interview, rebuttal, and counterfactual prompts; “red team” prompts to stress‑test bias.
• An evaluation rubric to score fidelity (content coverage, stance consistency, tone match, citation quality).
• A ready‑to‑run concept test (copy + images/video) and a debrief template with exec‑summary visuals.

Ethics, consent & guardrails

• Disclosure first: Always label outputs as simulations grounded in listed sources. No impersonation or endorsement claims.
• Consent where required: Use public figures or get consent for private individuals; avoid sensitive/PII content.
• Provenance matters: Keep a live bibliography; link every notable claim to a source.
• Scope the twin: Model opinions they’ve expressed; don’t invent hidden facts or private experiences.
• Bias checks: Compare twins across demographics/segments and run “challenge” prompts to surface blind spots.
• Human validation: Use twins to narrow options, then confirm with real participants before high‑stakes moves.

Proof points you can cite

• Cut concept‑iteration cycles from weeks to days by pre‑screening ideas.
• Improve stakeholder buy‑in with cited, side‑by‑side reactions from recognizable viewpoints.
• Reduce moderator time by rehearsing guides and hypotheses before fielding.
  

After the workshop, you’ll have

• Two usable twins (with disclosures and bibliographies)
• A persona set mapped to buying stages
• A tested prompt library and evaluation rubric
• A concept‑test template and debrief deck
• A simple governance doc (versioning, drift checks)

Meet Your Coach

Mark Michelson, founder of Threads Marketing Resaerch & Strategy and Southeast Regional Director for The AI Collective, is the lead coach and course creator for Stitchs.

Mark is a lifelong entrepreneur and connector who started and managed several research companies, trade associations, and is the founder/producer of CX Forums conferences.

Mark is a tech enthusiast and pioneer in using advanced tools, platforms, and devices for his research and conference businesses. He has been using AI tools since 2015, when Microsoft BI was first released.

Mark has extensive experience in a wide range of AI use cases including: quantitative and qualitative data analytics, marketing research, CX/UX design, vibe coding, creating agents, synthetic audiences, personas and digital twins.

As a UX researcher and trainer, Mark has developed a healthy habit of "pressure testing" new tech to ensure accurate and reliable results for his clients.