Faster Horses Won’t Save You
By Joana Diogo
- article
- AI
- Artificial Intelligence
- Advertising Experiments
- Concept Testing
Summarise with AI
This article is based on my experience attending the 2026 Corporate Research & Insights Summit – you can access the video replays and download the shared content here.
Corporate Research & Insights Summit 2026
In this fireside chat, Fiona Blades, founder of MESH Experience, talked with Phil Szymala, Head of Insights and Analytics Europe at Visa, about using AI to do research better. Phil welcomed speed gains in synthesis and storytelling, and warned that framing AI as faster and cheaper biases how organisations measure success.
His main example came from Visa’s Milano Cortina Winter Olympics sponsorship. A bold, K-pop-inspired AI-generated character aimed at Italian Gen Z left executives excited and nervous in equal measure. An AI-moderated live quant session with 100 Gen Zs gave it about 17% likability, with respondents calling it inauthentic and disconnected from Italian culture. Within a week, the team reworked it as a music platform for local producers and a vocalist, and a 1,000-person quant test lifted likability to about 80%. The campaign reached one in four Italian Gen Zs and briefly entered the international radio airplay top 100. Seeing those before-and-after numbers side by side was a real highlight!
Phil places AI mainly on low-stakes decisions, complementing human signals, and warned that trouble starts when it replaces rigour on high-stakes ones. He noted that vendors often price as if for high-stakes use. Other examples included social listening that deepened cross-border travel findings on currency anxiety and mistrust of banks, and AI pretesting of almost 500 past Visa ads to build a creative playbook. Synthetic audiences remain less proven, and he compared AI’s teething problems to the early days of online grocery substitutions.
Visa tracks AI usage alongside stories of better outcomes. His closing advice was to stop treating AI as one thing, and to experiment and publicise wins.
What Stood Out for Me
Better should rank above faster and cheaper.
Haste and cheapness both have well-known downsides, while almost everyone favours doing something better. It is a timely reminder that while speed is valuable, quality remains paramount in an era driven by fast results.
Fast AI signals can rescue a failing concept.
The quick AI read bought Visa a week to rework the idea before budget was committed to the wrong one.
Keep AI on low-stakes decisions – at least for now.
AI should enable rigour and complement human signals. Replacing rigour on high-stakes decisions is where teams get into trouble.
Scale unlocks analysis that was once unaffordable.
AI pretesting of almost 500 past Visa ads against signals such as memorability and brand equity produced a playbook of better-performing creative characteristics for briefs.
Implications for Insights Leaders
If our AI metrics only count speed and savings, we may miss what matters.
The success criteria we choose shape how the organisation uses AI. Putting better first changes which projects look successful and which ones we invest in next.
A few easy places to start:
- Add an outcome or decision-quality measure to every AI project’s success criteria.
- Rank criteria in project briefs in this order: better, faster, cheaper.
- Collect before-and-after stories showing where AI improved a real decision.
The line that stayed with me:
“If we take faster, you’ll get people who say, ‘More haste, less speed’. If you take cheaper, you’ll get people that say, ‘Buy cheap, buy twice’. But you rarely get somebody that says, ‘Don’t do anything better’.”
–Phil Szymala
