Why Your AI isn’t Working and How to Get AI Implementation Right

There is no single cause, but a consistent theme is that many AI initiatives start without a clear purpose.

One of the most important observations from the discussion was: “When you actually look at a number of the AI initiatives, there’s no clear value case.”

In practice, this often means organisations are implementing AI because they feel they should, rather than because they have defined a specific problem to solve. Board-level pressure to “have an AI strategy” can accelerate this, but without a clear link to business outcomes, projects quickly lose direction.

This creates a familiar pattern where technology is introduced first, and the use case is worked out afterwards. The result is fragmented activity rather than meaningful transformation.

Another key issue is how AI is positioned within the organisation.

One insight from the discussion was: “It shouldn’t be something that sat in IT or in the projects area of the business.”

Treating AI as a standalone project limits its impact. AI is not just a tool, it changes how work is done, how decisions are made, and how customer journeys are delivered.

This means AI implementation needs to be approached as an operating model shift rather than a technology deployment.

Without that broader view, organisations risk creating isolated solutions that do not connect to the wider business, making it difficult to scale or sustain value.

A recurring theme in the conversation is the risk of automating the wrong things.

One line captures this clearly: “We’ve got this great piece of technology… and nobody goes and asks the operation, what’s the problem you’re trying to solve?”

This is where many AI implementations begin to unravel. Instead of redesigning processes, organisations layer automation on top of existing inefficiencies.

The result is not transformation, but amplification of existing issues.

Another point reinforces this risk: “We’re just going to have broken processes AI’d.”

For contact centres, this often shows up as increased repeat contact, declining customer satisfaction, and more pressure on frontline teams who are left to deal with the consequences.

A common misconception is that automating responses is the same as solving problems.

As the discussion highlights: “A true outcome is did the customer absolutely get the resolution or the information they needed.”

This distinction is critical. AI can handle high volumes of interactions, but if those interactions do not fully resolve the customer’s need, the outcome is still poor.

In fact, poorly implemented automation can create additional friction, forcing customers to repeat information or move between channels.

This is reflected more broadly in customer service trends. UK Customer Satisfaction Index scores have shown long-term pressure on service quality across sectors (Source link needed), suggesting that efficiency gains have not always translated into better experiences.

For organisations, this means shifting the focus from deflection and cost reduction towards genuine resolution and experience.

One of the most practical examples discussed highlights how disconnected systems can undermine even well-intentioned automation.

Customers are often forced to interact with multiple systems, re-enter the same information, and navigate inconsistent processes. Each individual system may perform well in isolation, but the overall journey becomes fragmented and frustrating.

As one point in the discussion explains: “Everyone’s KPIs will be high… but actually at the end of it, the experience and your outcome still leaves quite a lot to be desired.”

This is a critical issue for organisations investing in AI. Without a joined-up view of the customer journey, improvements in one area can create problems elsewhere.

A key theme throughout the conversation is the need to think beyond individual use cases.

One insight puts it simply: “It’s that upstream and downstream knock-on impact… which I think is missing at the moment.”

AI implementations rarely operate in isolation. Automating one part of a journey can have unintended consequences elsewhere, whether that is increased demand, reduced quality, or new operational bottlenecks.

This is where many projects fail to scale. Even if a pilot shows promise, organisations struggle to extend it because the wider operating model has not been considered.

This reinforces the need for a more holistic approach, where processes, people, and technology are designed together rather than treated separately.

Many organisations successfully launch AI pilots but struggle to take the next step.

One of the reasons is a lack of clarity about what happens next. As highlighted in the discussion: “If it fails, what happens? If it delivers some value… what’s the next step? And if it does deliver… how do we scale that?”

Without clear criteria for success and a defined path forward, pilots can become isolated experiments rather than stepping stones to transformation.

There is also a tendency to accept partial success without addressing underlying issues. This leads to a growing backlog of improvements that never get prioritised, limiting long-term impact.

Moving beyond this requires upfront agreement on outcomes, success measures, and scaling plans, not just technical delivery.

AI is not a one-off implementation, it requires ongoing refinement.

As the discussion highlights: “It’s got to be a continuous improvement loop.”

This means organisations need to go beyond high-level metrics and actively gather feedback from customers and employees. Understanding where journeys break down, where customers drop out, and where experiences fall short is essential to improving outcomes over time.

It also requires curiosity. Another key point raised was the importance of continuing to listen and learn, rather than assuming the problem has been solved once automation is in place.

For contact centres, this creates an opportunity to use interaction data, analytics, and AI itself to continuously improve performance and customer experience.

While the challenges are significant, the discussion ends with a clear and practical message.

One of the final points captures it well: “Be really clear around the purpose and what you’re trying to achieve.”

Clarity of purpose remains the most important factor in determining success. Organisations that define clear outcomes, understand their customer journeys, and align AI implementation with business strategy are far more likely to see value.

There is also a strong case for learning from others. Many organisations are already delivering successful AI use cases, and understanding what works in practice can accelerate progress.

Ultimately, the goal is not to implement AI for its own sake, but to use it as a tool to improve outcomes, reduce friction, and create more effective and resilient customer operations.

For CX and contact centre leaders, that means stepping back from the technology and focusing first on the problem.