‍

On 2 September, Arlo, Electric Twin and Focaldata brought together Insight leaders, brand and marketing teams, and AI/ML builders at Atomico's London offices for an evening built around one question: as AI takes on more of the execution in research, what's left for the Insight function to do?

The fireside chat paired two companies that have made opposite bets on what gets automated. Arlo, spun out of Focaldata, keeps the respondent and removes the researcher's manual labour – briefing specialist agents the way you'd brief a researcher, then judging what they deliver. Electric Twin goes the other way: synthetic audiences built from a company's own customer data, standing in for the respondent while keeping the researcher's judgement in the loop.

Justin Ibbett (Founder & CEO, Focaldata) and Ben Warner (Co-Founder & Chief Data Scientist, Electric Twin) opened each theme before handing the room the conversation. Discussion covered:

• Is this shift different in kind, or just bigger than the last few? AI and synthetic data get talked about as a bigger disruption than panels, mobile or DIY research before them — the question is whether that's really true, or whether it just feels that way because it's happening to this generation of the industry.
• Two opposite bets on what's disposable. Arlo removes the researcher's labour and keeps the respondent; Electric Twin removes the respondent and keeps the researcher's judgement. Whether both bets can be right at once, or whether the market resolves toward one of them, was a central thread.
• Trust as the real bottleneck. Model capability isn't what's slowing adoption down — knowing when a client can act on an answer versus when it needs escalating to a human is. The conversation pushed on what a concrete test for that actually looks like in practice.
• Where the researcher's role goes next. As execution automates, the job shifts up the stack: less running projects, more setting standards, calibrating the system, and handling the judgement calls the system still can't make. That has direct implications for how research teams get resourced and structured.
• Throughput vs. transformation. Sold as cheaper, faster research, this commoditises the category and prices only move one way. Sold as genuine transformation, decisions move to different people with different budgets, and today's Insight team looks smaller and different. Which future either company is actually building toward — and what winning looks like in each — closed out the evening.