Truth Tips: Learnings from a Synthetic Data Pilot

A lot is happening in the world of AI right now. But not all developments are equally transformative. At Truth Consulting, our approach to innovation with AI is to stay grounded. That means testing, piloting, and learning—always in close partnership with our clients. We’re not here to chase shiny objects; we’re here to create tools that work for our clients in the real world.

For the past six months, we've been working on a pilot to create synthetic personas—also known as digital twins—using AI to support the activation of a 28-market personal care segmentation. Being able to test and generate new concepts with specific segments could be a huge advantage for our clients – even transformative!

But naturally, this raised some big questions: Can AI generate personas that feel human? What data works best—qual, quant, new, old? Which tools and platforms are genuinely useful?

This edition of Truth Tips shares what Victor Reynoso and Carlota Rosa-Delgado have learned while creating solutions with the latest cutting-edge tools and techniques in the market:

1. Focus on commercial value over gimmicks. Just because AI can do something doesn't mean it should. Always ask yourself what the specific use case is and how what you’re building will drive commercial value. Is it creating sharper creative briefs, targeted innovation, better CX journey mapping, or growth opportunities?

   
   

2. Be sceptical of big promises. If it sounds too good to be true, it probably is. Tech vendors may promote features that aren't live yet. Always test before you commit and inquire about their roadmap for upcoming features—and when they’ll actually arrive.

   
   

3. Trial-and-error loops are essential. The only way to know what works is to try platforms yourself: What data can be used to train algorithms? Can your data be ingested and analysed? How much data cleaning is required? Is the interface user-friendly? How useful are the answers for idea generation?

   
   

4. It’s all about the pilot – start small and iterate. Your first pilot version won't be perfect—and that's okay. Instead, build a minimum viable offer (MVO) with limited scope and clear metrics—e.g. speed of insight, idea quality—then refine your approach before scaling.

   
   

5. Simple is often best. A platform full of features might excite AI evangelists, but your client—especially one with multiple non-research stakeholders—needs something intuitive and easy to use. Sophisticated solutions or platforms are useless if client teams find them hard to access, query, or integrate into their workflows.

   
   

6. Test against your gold standard. If you have the opportunity, run a test alongside your primary research methodology to assess accuracy. This helps build buy-in and trust in the approach with stakeholders.

   
   

7. Keep it fresh and futures focused. Don’t rely solely on historical inputs. Use tools like social listening, futures work, and semiotics to keep segments alive and culturally relevant.

   
   

8. Partnerships matter. Working openly with clients and suppliers is essential to building AI solutions with integrity and long-term value. As AI evolves, track what works, what doesn’t, and share those insights—it’s valuable in itself.

If you have a brief that you think could be tackled with generative AI—and you want to assess its commercial potential—get in touch: hello@truth.ms