Food Matters Live Podcast · Food Matters Live

AI's impact on NPD, with Antonio Weiss, Author and Tech Expert

·34 min·2 clips

"Don't use ChatGPT to do reformulation": Weiss separates large language models from specialist graph neural nets in food R&D.

1. Food Matters Live Podcast recorded this session live at the Food Matters Live event in Dublin in November 2025, asking whether AI will revolutionise new product development (NPD) in food or is mostly hype. 2. Host Anthony Zahra interviews Antonio Weiss, author of AI Demystified for the Financial Times and a digital transformation partner who began working in AI in 2018 when the UK government set up the Office for AI. 3. Weiss's core thesis is that AI is a general purpose technology like the smartphone, the internet and the steam engine, and that anyone claiming to know how it will work in two or three years is "wrong". 4. He separates large language models such as Gemini, ChatGPT and DeepSeek, useful for marketing, naming and trends, from specialist graph neural nets and machine learning models needed for flavour prediction, demand forecasting and reformulation. 5. He cites Coke and PepsiCo using social listening, web scraping and sentiment analysis for Gen A trends, and Carlsberg speeding up R&D flavour prediction by adding nutrient, ELN and focus group data to an open source GNN model. 6. On creativity, Weiss sets an optimist scenario, where taste is "everything" in Rick Rubin's phrasing and humans choose between 3 to 10 formulations, against a pessimist one where investors' return comes from replacing jobs, as ready meals and Uber Eats displaced earlier habits. 7. He describes agentic AI as autonomous strings of tasks, such as creating a video ad, converting it into Facebook, X and TikTok posts and geo-targeting customers, and says it could push marketers toward selling to consumers' AI agents within three to five years. 8. In audience Q&A, he says AI search has an environmental cost, comparing an early ChatGPT query to boiling a kettle versus a light switch for Google and noting data centres in drought-hit Querétaro, Mexico. 9. He advises starting small, avoiding blank-sheet prompts like "come up with a new product tomorrow", and measuring performance before and after AI, since roughly 85% of digital transformation efforts fail and an MIT study put AI failure at 95%. 10. Suited to food and drink professionals weighing practical AI adoption.

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from Food Matters Live Podcast