Can AI simulate humans?
We spent three months finding out.
40+
interviews across vendors, brands, and academics
16
vendor platforms, self-serve to enterprise builds
13
brand-side insights teams piloting, deployed, or building
✺
Independent + unsponsored
After reading the report, you’ll be able to:
- 01
Understand the four approaches and where each one fits
- 02
Cut through inconsistent vendor terminology and claims
- 03
Match the right approach to the right research need
- 04
Ask better questions in vendor conversations
- 05
Know what exercise to run to validate a vendor before you rely on it
How it works
The four approaches
How to evaluate and buy
How can we put this in practice
Simulated Humans Foundations
Sized for a team. Larger audiences work for session one.
Who it's for
Insights teams who will run, review, or sign off on simulated studies this year. Also I&A forums, marketing summits, and leadership offsites.
What the two hours cover
What your team walks out with
A shared vocabulary so your team, your stakeholders, and your vendors are describing the same thing.
The vendor question checklist covering grounding data, reasoning, validation, drift, and data rights.
A prompt-only starter kit your team can run in Copilot, ChatGPT, or Claude the following week.
From Use Cases to a Pilot Plan
Who it's for
The VP of Insights with their team members and/or cross-functional stakeholders from marketing, analytics, or innovation.
What we cover
Your business decisions first. Which decisions this needs to support in the next year, and for which audiences.
Outline the design considerations for each persona group and use case, along with data requirements and gaps.
How you'll validate. What "validated" means for your team and your stakeholders before anyone runs a comparison.
After workshop deliverable
Our recommended technical approach by business need.
A short list of vendors that match the technical approach, with a build option.
A vendor test brief ready to use with vendors.
04 · Pilot Advisory
Vendor Evaluation and Pilot Design
Scoped to your use cases and the number of vendors in play.
Who it's for
Insights leaders with budget approved to run pilots on two or more vendors.
What we cover
Defining the best test benchmark to pressure test the vendors.
Define validation criteria.
A pilot test design to make a clear vendor recommendation and implementation plan.
What you leave with
A scored vendor recommendation with the reasoning shown.
A pilot plan with success criteria.
A checklist needed to ensure legal and procurement are aligned.
05 · Custom Consulting
Bring Us the Problem
A marketing team has already built a persona tool and you need someone to audit it. A supplier needs to decide where it sits in this market. Two functions are fighting over who owns the vendor decision. We do custom consulting.
Every custom engagement starts with a free 30-minute consultation call and oftentimes requires another free 30-minute follow-up scoping call before a custom proposal is built.
Who it's for
Insights leaders with a specific situation on their hands, and research suppliers rethinking their offer for a market where their clients can simulate the study.
Work we've been asked for so far
Internal build audit. Someone in your company already built a simulation tool. We review the grounding data, the prompts, the scoring, and the UX, and tell you what it can and can't be trusted for.
Build your own solution. Want to create your own solution with the data you have but lack the technical ability? We have the team and the talent.
Supplier repositioning. Where your agency or platform fits on the market map, what your clients will be asking you next year, and what to build or partner for.
Ongoing advisory. A standing monthly slot for the leader who wants a second opinion on every vendor call, pilot readout, and board question.
Your consulting dream team
Kristi Zuhlke
Technology and brand-side insights
Started in insights at Procter & Gamble, then founded and sold the technology company KnowledgeHound. Now advises Fortune 500 insights leaders on where AI fits in their work.
Remy Thim
Market research execution
A career on the supplier side running rigorous methodologies and designing new approaches. Has vetted AI research tools hands-on, as a practitioner.
Matt Macbeth
Building models that simulate humans
Deep technical expertise in how simulation systems are built, having built his own systems that run on off-grid supercomputers.
Dr. Tom Foust
Linguistics
Expert in how language models interpret and produce human language, which decides whether a simulation holds up.
Get started
Get the report today.
An independent, unsponsored guide to simulated humans in market research. Read it on your own, or bring us in to help put it into practice.
- The full 15-section report in PDF format
- An explanation of how this methodology actually works
- A map of the four approaches
- A guide to run a validation study
- Built on 40+ interviews, 16 vendors and 13 brand teams
Consulting solutions
Report includedFrom first read to a working pilot.
- 01 UnderstandThe full report, so everyone starts from the same evidence
- 02 AlignTeam training on how to judge the method and its vendors
- 03 DecideA workshop that maps the framework onto your own studies
- 04 TestPilot advisory while you put a vendor to work on a real decision
- 05 AdoptCustom consulting to build the method into how your team works
- 06 You Tell UsTell us your biggest challenge and we will create a custom proposal
Every engagement is scoped and priced to your team.
Which vendor does the report recommend?
None. Nobody paid to be in it. The report gives you the framework and skills you need to assess the vendors you are currently looking at.
When you say "simulated humans," is that synthetic?
Many people use the word synthetic to describe what we call simulating humans. We believe the word synthetic implies fake or inferior and gives people the wrong idea of what the method does. Simulating humans is the most accurate descriptor of what's happening.
Does this report cover sample augmentation?
No. The report covers replacing a study you'd otherwise field with people. Augmenting a sample with simulated responses is a different methodology and technique and sits outside the scope of this work.
I'm a research supplier. Is this for me?
Yes. We work with agencies exploring this as an offering of their own. Several of the offerings above are adjusted for the perspective market research suppliers bring when evaluating solutions they could build or bring on themselves.
