The technology can already do 95% of what your business needs from AI. So why are 70% of AI projects still failing?
Asa Cox reckons the tech's already good for 95% of what your business needs - so the problem was never the AI, it's everything wrapped around it.
Key takeaway most AI projects don't fail on capability, they fail on execution.
Speaking at the 2026 AI Roadshow in Auckland, Asa Cox - founder of Pneuma - puts a pointed question to the room: if the technology can already do 95% of what a business needs from AI, why are 70% of AI projects still failing? His argument is that the bottleneck has shifted. The capability is more than ready; what stalls projects is everything around it.
For a New Zealand audience weighing their first serious AI investment, it's a timely reframing. Cox's point cuts against the instinct to keep waiting for the tech to mature - the tools are here, and the harder work is organisational: picking the right problem, wiring AI into real workflows, and seeing a project through past the pilot.
It's a grounded, un-hyped message for the Roadshow crowd - less about the next model and more about why so many good intentions never make it to production, and what it takes to land in the 30% that do.
“Turns out the AI was ready all along - it was the rest of us still finding our feet.” — The Veteran Critic, The 6 O’Clock News