Frequently asked questions

No. Persona Dynamics focuses on grounded outputs linked to your evidence base. Rather than generating responses from generic model training alone, it keeps responses tied to the research and source material you provide. For a fuller comparison, read AI personas vs synthetic users.

Typical inputs include customer and audience research, survey findings, reports, internal documents, strategic materials, and specialist domain knowledge. If content is relevant evidence for the use case, it can usually be included.

No, and it tends to do the opposite. Persona Dynamics increases the return on research you have already paid for: work that would otherwise be filed and forgotten becomes a living evidence base your team can question again and again. When existing research visibly shapes decisions, the case for commissioning more of it gets stronger, not weaker. Human judgement stays central, and every answer remains traceable to its source.

It is for teams that want to close the loop between research and the decisions it should inform: strategy, research, product, service design, marketing and specialist-domain teams. The aim is to make existing evidence quicker to reach and easier to act on, so it shapes decisions while it still matters. The AI is the means, not the point: what matters is research that is traceable, accountable and actually used.

Yes. Delivery can be platform-only, guided setup, or a managed engagement. The right model depends on your internal capability, timeline and use case complexity.

Start with a short conversation about your goals, evidence sources and intended users. From there we can recommend an initial scope and the most suitable delivery model.

Synthetic respondents are AI systems that stand in for human participants and generate answers to research questions. How they work varies by tool. Some produce responses from a general language model's training data, which can be plausible but ungrounded. Persona Dynamics® takes a different route: its Dynamic Personas™ answer through retrieval-augmented generation over your own primary research, so each response is drawn from named evidence in your knowledge base and linked back to its source passages.

It depends on the tool, and the distinction matters for any decision you base on the output. Many AI personas answer purely from a general language model, producing fluent but ungrounded responses. Persona Dynamics® personas are built from your organisation's own primary research and answer through retrieval-augmented generation, so each response draws on specific evidence in your knowledge base. You can tell which kind you are using by checking whether the persona shows the source passages behind every answer.

Accuracy varies by tool and should be measured, not assumed. In an independent, blind evaluation, Persona Dynamics® personas scored 92.4% accuracy on core claims, marginally above the human-expert benchmark in the same study. Accuracy is necessary but not sufficient: Persona Dynamics triggered Critical Failure flags more often than the human experts, which is why human review is mandatory in every deployment and no output is treated as determinative on its own.

Ask for an independent, blind evaluation against a human benchmark, not a single platform-wide accuracy figure. A credible benchmark compares the tool's answers with expert responses to the same questions, scored by assessors who do not know which is which, and it discloses failure rates as well as successes. Persona Dynamics® ran exactly this design, with 9 independent subject matter experts scoring responses blind, and reports its Critical Failure rate openly rather than only its headline accuracy.

Accuracy is only meaningful relative to a benchmark, and human experts are an imperfect one. In Persona Dynamics® validation the human experts themselves triggered Critical Failure flags in 40% of responses, so accuracy here means performance measured against real expert answers under identical, blind conditions, not against a flawless ideal. This is why Persona Dynamics reports comparative results rather than an absolute score, applies confidence scoring to each answer, and keeps human review in every deployment.

Yes. With Persona Dynamics® every answer is traceable to the evidence behind it. Each response is accompanied by a source evidence panel showing the specific passages from your research that informed it, and the Digital Thread links the output back to its underlying material. This means you can inspect, quote and challenge the basis of any answer rather than taking it on trust, which is the central reason Persona Dynamics is built around retrieval from your own knowledge base.

Persona Dynamics® treats confident errors as a known risk and is designed to surface them. In independent validation it triggered Critical Failure flags in 51.2% of responses, more often than human experts, which is exactly why every answer carries a confidence score and grounding star ratings, and why human review is mandatory. To judge which answers to trust, check the confidence indicator and the source passages, and escalate low-confidence or unsupported outputs to a person.

Sycophancy is a real risk with synthetic users that answer from a general language model, since they tend to produce agreeable, plausible responses. Persona Dynamics® reduces this by grounding every answer in your actual research through retrieval-augmented generation, so a Dynamic Persona™ reflects what the evidence says rather than what sounds encouraging. Where the research contains criticism, reluctance or contradiction, that comes through, and the source passages let you see whether an answer is supported or merely positive.

Grounding reduces hallucination but does not eliminate it, so Persona Dynamics® is built to make it visible. Answers are generated by retrieval from your own research rather than open-ended generation, and hallucinated content is one of the explicit Critical Failure flags measured in validation. Because no system is immune, each answer carries a confidence score and source passages you can check, and human review is mandatory in every deployment so unsupported claims are caught before they are acted on.

No. Persona Dynamics® does use large language models, but its answers come from retrieval-augmented generation over your organisation's own research held in a siloed knowledge base, not from the model's general training data. Each response is linked to the specific source passages behind it through the Digital Thread, carries a confidence score, and has been tested in independent, blind validation against human experts. A general chatbot offers none of this evidence grounding, traceability or measured accuracy.

Persona Dynamics® personas do not invent facts beyond your evidence, but they surface what is latent within it. By retrieving and synthesising across an entire research corpus, a Dynamic Persona™ can connect findings that sit in separate documents, expose patterns no one had queried, and bring contradictions to the surface, all linked to the source passages. The new value is in making a large body of existing research continuously interrogable, not in generating claims your evidence cannot support.

The substance of an answer stays consistent because Persona Dynamics® retrieves from the same fixed evidence in your knowledge base each time, and the source passages let you confirm what any answer drew on. The wording can vary between runs, as it does with any language model, but the underlying evidence does not. To guard against drift after updates, the Golden questions framework re-runs a fixed set of benchmark questions and flags changes for review before deployment.

Use synthetic users when you have little or no research and the stakes are low, for fast early-stage ideation and hypothesis generation. Recruit real participants when you need new evidence about how people behave in practice, especially for consequential decisions. Persona Dynamics® sits in a third position: it makes your existing primary research continuously queryable, as evidence-grounded decision support with traceable sources, rather than substituting for either synthetic stimulus or direct engagement with real people. The full breakdown is in AI personas vs synthetic users.

Persona Dynamics® stores each organisation's data in a separate, siloed vector store on UK-hosted AWS infrastructure, kept logically isolated from other clients' data. Your research is never used to train any AI model, ours or a third party's, and data is deleted on account closure. Persona Dynamics personnel do not access your research content directly, since processing is automated, and data processing agreements are available for larger clients and managed engagements.