Why Most AI Projects Collapse And How To Build One That Works

AI projects are failing at a rate that should make any leadership team pause, and the uncomfortable truth is that the model is rarely the real problem. We sit down with Andy Hayler, president and CEO of the Software Industry Authority and a long time data strategist, to explain why organisations keep shipping AI initiatives that look impressive but deliver zero measurable value.

At A Glance / TLDR:

• the 95% AI project failure rate and what it signals
• poor data quality as the top cited cause of failure
• why LLMs are probabilistic and why hallucinations are inevitable
• real world examples of confident errors in legal and strategy work
• choosing the right type of AI for the job, not defaulting to LLMs
• what the top 5% do differently: ownership, governance, ROI, measurement
• building AI literacy so teams know limitations and safe use patterns
• starting small with high value use cases and scaling via proven wins

We dig into the biggest repeat offender: data quality. When teams bolt generative AI onto messy corporate documents through retrieval augmented generation (RAG), the system can only reflect the gaps, contradictions, and missing ownership already baked into the data estate. From there, we tackle the misconception that large language models behave like normal software. LLMs are probabilistic token predictors, not deterministic calculators, which is why hallucinations and “confidently wrong” answers show up in high stakes areas like law, medicine, and engineering unless you design proper human review and verification.

Andy also breaks down the “AI is one thing” myth, contrasting LLMs with machine learning for predictive maintenance and reinforcement learning breakthroughs like protein folding. The practical takeaway is an operating model: pick the right technique, define success with ROI and clear metrics, assign business ownership for data, and start small so early wins build confidence and capability across the organisation.

If you want to be in the 5% that succeed, subscribe, share this with a colleague who owns delivery, and leave a review. 

Where is your organisation most likely to be “confidently wrong” with AI?

LinkedIn: Andy Hayler

Andy's Book on Amazon UK: Beyond The Hyper: A Realists Guide to AI

Support the show

𝗖𝗼𝗻𝘁𝗮𝗰𝘁 my team and I to get business results, not excuses.

☎️ https://calendly.com/kierangilmurray/results-not-excuses
✉️ [email protected]
🌍 www.KieranGilmurray.com
📘 Kieran Gilmurray | LinkedIn
🦉 X / Twitter: https://twitter.com/KieranGilmurray
📽 YouTube: https://www.youtube.com/@KieranGilmurray

📕 Want to learn more about agentic AI then read my new book on Agentic AI and the Future of Work https://tinyurl.com/MyBooksOnAmazonUK