It’s not the AI’s fault, guest blog by Zudu CEO Paul Duffy

Every boardroom conversation seems to land on AI right now, from “we need it” to “be careful with it,”  to “why don’t we have it yet?”  Budgets are being allocated, pilots seem to be underway, and innovation teams are busy.

The problem is that most of that activity isn't translating into real outcomes. The easy explanation is that AI is “still early.” It sounds reasonable, and it buys time, but it is not accurate. The models work. The tools are accessible. In most cases, the limitation is not the technology - it is the business that the AI is being dropped into.

What I see time and time again, is organisations expecting AI to compensate for years of operational compromise. Systems that have been layered on top of each other without much thought for how they connect. Data is spread across platforms that were never designed to work together. Processes that rely on people manually stitching everything together just to keep things moving.

Most businesses have a small group of people who hold that together. They know where everything is, how it fits, and how to fix it when it breaks. They are essential, but they are also a sign that the operation is more fragile than it should be.  AI does not solve that.

Even if you introduce very capable models into that kind of environment, you tend to get the same outcome. Early excitement, a promising proof of concept, and then a slow drift as the reality of integrating it properly sets in. When it stalls, the technology often gets the blame.  In reality, it is doing exactly what it should. It just has nothing solid to work with.

The businesses seeing genuine returns from AI are not starting with tools. They are starting with how the business actually runs. Where time is being lost, or where decisions are slowed, or where people are doing work that should not exist in the first place.

Once those problems are clear, AI becomes useful. Before that, it is just another layer on top of an already stretched system.  Turning that into something that works in a live environment is not straightforward. It means connecting to production systems, structuring data properly, and changing how teams operate day-to-day. That work is not particularly visible, but it is where the value sits.

There is a gap between what the technology can do and what most organisations are set up to support. Closing that gap is a delivery challenge, not a technology one.

So the question for leadership teams is not whether they are doing enough AI. It is whether their business is ready to make use of it, and whether they have the capability to turn potential into something that actually operates at scale.

That is a harder conversation than approving another AI pilot.  But crucially, it is the hard conversations that lead to the most beneficial results.