Frequently asked questions

FAQ

• What is a digital twin?

  A Brox digital twin is a 1:1 behavioural replica of one real, verified person, built from hours of AI-moderated video interview. It predicts how that specific person would actually decide, rather than describing an average or a persona. In aggregate, a panel of twins lets you test ideas against real human behaviour without running a new study each time.

• How accurate are Brox digital twins?

  Brox twins predict individual real people's decisions with approximately 85% accuracy at the individual level, measured against the real people each twin is based on. Accuracy is strongest where it has been validated against real-world behaviour, such as financial decisions. We lead with the individual-level figure because it is the hardest and most useful bar to clear.

• How is a Brox digital twin different from an LLM or a ChatGPT persona?

  A general-purpose LLM is trained on text, not on the behaviour of specific people, so when you ask it to role-play someone it reaches only around 48% behavioural accuracy and averages the differences between real people into a single voice. A Brox twin is built from a real person's interview, reaches approximately 85% accuracy at the individual level, and can always show you why it answered the way it did. An LLM cannot tell you why it said what it said; a digital twin always can.

• How are Brox digital twins built?

  We recruit and verify real participants, then run roughly five hours of AI-moderated video interview per person. From that we build a 1:1 twin, which is then continuously enriched with follow-up pulses over time. Every answer a twin gives traces back to that source material.

• What is Ask Brox?

  Ask Brox is the self-service platform for running natural-language research against your twins. You ask a question in plain language and get a dashboard with metrics and Decision Drivers in under 15 minutes, with no research team required. For larger or more complex work, Brox can also run fully managed engagements.

• What are Decision Drivers?

  Decision Drivers are the traced reasons behind a prediction: the why behind what a twin says, linked back to the real interviews the twin was built from. They are what let you audit an answer and sense-check it down to the interview moment, instead of trusting a black box.

• How fast do I get results?

  Self-service questions in Ask Brox return dashboards with metrics and Decision Drivers in under 15 minutes. More complex, fully managed simulations are typically delivered in roughly 24 to 72 hours. Either way it is hours, not the weeks a traditional research cycle takes.

• How is Brox different from traditional market research?

  Traditional research takes weeks, costs a lot, and ends with a report that goes stale. Brox lets you run unlimited questions against a panel you can re-query indefinitely, with answers in minutes to hours and every result traceable to real people. The panel is a durable, reusable resource rather than a one-off study.

• What's the difference between digital twins, synthetic personas, and synthetic twins?

  A synthetic persona is an archetype averaged from survey data; a synthetic twin is a character a model invents with no real-person grounding. A Brox digital twin is a behavioural replica of one specific real person, built from hours of interview and traceable back to them. Only the digital twin preserves the variance between real people and can be validated against ground truth.

• Can I build a custom panel of a specific audience, such as doctors or high-net-worth individuals?

  Yes. Alongside everyday-consumer panels, Brox builds custom panels of hard-to-reach specialists such as healthcare professionals and high-net-worth individuals (for example those with $5M+ investable assets). We recruit, interview, and build twins for the exact audience you need, then let you query them indefinitely.

• What industries and use cases is Brox used for?

  Brox is used wherever a decision depends on predicting what real people will do: financial services, healthcare and pharma, media and entertainment, sports, and consumer brands, among others. Typical use cases include concept and message testing, pricing and price sensitivity, segmentation, switching intent, and complex scenario simulation. It is the human behaviour layer for every decision system.

• Is my data secure and is Brox compliant?

  Yes. Our SOC 2 Type II audit is complete, with the report available to clients under NDA once issued. SOC 2 Type II is the standard that verifies security controls are not just designed but operating effectively over time. Participants are real, verified people who consent to taking part.

• How many digital twins does Brox have, and where?

  Brox has 30,000+ active digital twins across the US, UK, Japan and Turkey, with more markets planned in 2026. They are used by around 20 enterprise clients.

• Do twins stay current over time?

  Yes. Twins are continuously enriched with follow-up pulses after the initial interview, so they stay current as people's views and circumstances change. That is part of why a twin panel appreciates as an asset rather than going stale like a one-off study.

• What questions does Brox not answer well?

  Brox is built to predict the behaviour of real people, so it is not the right tool for tasks that have nothing to do with human decisions, such as drafting copy, summarising a meeting, or debugging code, where a general-purpose LLM is better. It is also most reliable for audiences we have actually interviewed, and we would rather build the right panel than over-claim on one we do not yet cover. Where behaviour has not been validated, we say so.

• Who founded Brox.AI and when?

  Brox.AI was founded in 2023 by Hamish Brocklebank (CEO, based in San Francisco) and Durgé Seerden (CPTO, based in Amsterdam).

• Why not just build my own LLM?

  You can prompt your own LLM with a segment profile and get a confident, plausible, cheap answer, but it will be a hallucination dressed up as evidence, and no one will be able to tell you why it said what it said. LLMs are trained on what people wrote in public, not on how they privately decide, and impersonation averages real people into a single voice. For any decision where you need to predict real human behaviour with an audit trail, a twin panel is the right instrument and a prompted LLM is the wrong one.