Why 95% of enterprise AI projects fail:
Expertise, not technology, is the key

The mirage of plug-and-play AI

AI vendors make it sound effortless: plug in a large language model, flip a few switches, and watch your customer service transform overnight. The truth is more sobering. According to a recent MIT study, 95% of enterprise AI pilots fail to deliver measurable ROI.

At this year’s UK National Contact Centre Conference, Stuart Dorman, Chief Innovation Officer at Sabio Group, delivered an even more uncomfortable truth: technology isn’t the problem — expertise is. “Big tech wants you to believe AI deployment is simple: just plug in and transform your operation,” Dorman told delegates. “The reality is starkly different. Anyone can build a demo bot in an afternoon, but scaling AI to handle thousands of voice interactions at high performance levels requires expertise that many organizations simply don’t possess.”

The expertise deficit – the real bottleneck in AI

While model costs have plummeted, OpenAI’s token prices have dropped more than 1,000-fold in three years and the specialist skills needed to exploit them are increasingly scarce. Dorman calls this the AI paradox: as technology costs fall, implementation costs rise.

Success requires a rare blend of UX design, linguistics, prompt engineering, data science, systems integration, and contact center operations. In Dorman’s words: “The expertise to exploit this technology is increasingly difficult to find. Demand is outstripping supply.”

For many organizations, this means their pilots never escape the lab. They lack the multidisciplinary orchestration needed to make AI work in production.

“The organizations thriving in AI transformation aren’t those with the biggest budgets,” Dorman concluded. “They’re the ones who recognize that expert guidance through the complexity of real-world deployment is...