Most companies are failing at operations and blaming AI. It’s not the AI’s fault.
Our CEO, Paul Duffy, gives us his latest thoughts on why businesses are blaming AI.
The easy explanation is that AI is “still early.”
Every boardroom conversation inevitably lands on AI right now. Budgets seem to be shifting, AI pilots are running, and innovation teams are well and truly underway. Businesses are busy, but they’re not producing the outcomes.
The easy explanation is that AI is “still early.” It lets everyone off the hook and isn’t accurate. Most models work, and most of the tools are accessible. If your AI programme isn’t delivering, the technology isn’t where the problem lies.
What I keep seeing is organisations with years of accumulated operational debt expecting AI to paper over it. Core systems bolted together over time. Data scattered across multiple platforms that were never meant to connect. Processes that only function because a group of absolute legends are manually bridging the gaps, often the A-team in the office. You know the group, all aged 30-40, who look 60-70, each storing 180 passwords on the back of a fag packet, with expert knowledge of columns in a spreadsheet saved on their desktop. (sound like you? Call me.)
AI doesn’t fix that environment. Drop a capable model into a fragmented operational landscape, and you get exactly what you’d expect. An impressive demo, a stalled rollout, and a post-mortem that blames the tech. The tech was fine. The foundations just weren’t ready for it.
Businesses that see real returns from AI don’t start with a technology decision.
They start by asking where the business is genuinely slow. Where manual intervention is the norm. Where people spend their time moving information between systems rather than doing anything useful with it. Those are operations problems. AI becomes the solution once the problem is clearly defined.
Getting from that diagnosis to working software inside live systems takes serious delivery capability. Connecting AI to production environments, structuring data so it’s actually usable and embedding new processes into how teams work day to day. That work isn’t glamorous, and it takes time. It’s also the difference between a pilot and something that works long term.
The genuine question for most leadership teams isn’t whether they’re doing enough AI. It’s about whether the operational foundations are in place to support it and whether they have the delivery capability to close the gap between what the technology can do and what the business currently allows it to do.
That conversation is less comfortable than approving another innovation budget. It tends to be more useful.
