How to evaluate a synthetic research vendor: 11 questions for professional services firms to ask

Synthetic personas and digital twins are convincing, and that's the problem. A weak synthetic research vendor's output looks as convincing as a strong one's, so you can't judge the work by the deliverable. You judge the method. This field guide gives law firms, accounting firms and other professional services firms eleven questions to put to any synthetic research vendor, and what strong and weak answers sound like. It draws on Park et al., LLM Agents Grounded in Self-Reports Enable General-Purpose Simulation of Individuals (Stanford and Google DeepMind; arXiv, version 3, June 2026): https://arxiv.org/abs/2411.10109 The written guide, with every question: https://taleist.agency/asi/evaluate-s... How synthetic research providers differ: https://taleist.agency/asi/synthetic-... Book a 30-minute call: https://taleist.agency/asi/#book The narration is an AI clone of Steven Lewis's voice, made with ElevenLabs. 0:00 How to judge a synthetic research vendor 0:56 What synthetic research means: personas and digital twins 1:44 Where synthetic personas go wrong: AI model bias and the average 4:48 Two questions to ask any synthetic research vendor 6:42 Testing the method: AI models, scores and benchmarks 9:12 Testing what synthetic personas tell you 11:16 How to read a vendor's answers 12:09 How Taleist does synthetic research for professional services