Ben Warner

Co-founder, Electric Twin (AI "synthetic audiences") · Expert Watch · weekly scan

Why we watch him. Technical founder of one of the most-funded UK startups in the "simulate humans with LLMs" category (Atomico-led round, Feb 2026), and a government-tested modeller (Vote Leave 2016, the Conservatives' 2019 model, SAGE during COVID). A live, articulate voice on where AI simulation of opinion genuinely works versus where it is hype. (His brother Marc Warner runs Faculty AI; not to be confused with US Senator Mark Warner.)

Current headline view: AI synthetic audiences are a new, complementary tool that adds speed and scale to understanding people, not a crystal ball and not yet a replacement for talking to real humans.

Latest 1-page summary (as at 9 Jul 2026)

  1. Complement, not replacement. "Synthetic audiences are an additional tool in most cases, rather than a replacement" (a new saw alongside the hammer and screwdriver).
  2. Not a crystal ball. Concedes synthetic sampling "is not a crystal ball", pushing back on over-claiming.
  3. Traditional research is too slow and expensive. Frames the incumbent model as "six weeks and £50,000" per project.
  4. The say-do gap justifies new methods. "People do not purchase what they say they will", so self-reported survey data is not a gold standard.
  5. Methods must keep evolving. Research should "iterate, always evolve, always improve".
  6. Beware "innovation theatre". Do not deploy AI where removing it would not change the user's work; value over novelty.
  7. Whole-hearted adoption, with humility. Uses AI "for pretty much everything"; "it's really quite scary to think this is the least powerful they'll ever be."
  8. Accuracy claim (watch critically). Electric Twin claims its synthetic audiences match real people ~96% of the time (LSE validation cited). Treat as a company claim, not settled fact.

Notable reactions (operators & researchers)

High-signal reactions from operators or researchers (not commentators) to a specific argument of his. Challenges first. As at 9 Jul 2026 (baseline; entry below predates the standard weekly window and is included as the defining recent challenge).

Quick links

LinkedIn Electric Twin News