Chasing AI without a clear use case

Right now, we have an ‘AI gold rush.’ Whether in the boardroom or on the front line, everyone wants a piece of the action. But many projects are doomed before they begin because they skip the most important questions: do we have a clear understanding of our business challenges? Do we understand and can we define a specific use case? Simply put, what are we trying to solve?

Too often, AI is marketed as some kind of technical panacea, when, in reality, it’s ultimately just a capability, of itself – not a solution. I’ve seen organisations invest heavily in the platforms and technologies – highly polished, AI-powered chat agents, for example – without ensuring they have a clear understanding of the challenge they are trying to solve, a well-defined adoption strategy, the internal resources, or data content and structures needed. The result? Frustrated consumers, low adoption rates and brand damage.

Success initially comes when CIOs start small and have very specific outcomes defined. If you pick a challenge that’s repetitive and measurable, like invoice processing or data aggregation, you can very quickly prove the value with simple automation. You can then expand into more sophisticated use cases; whether that is in the Copilot space for collaboration and generative AI use cases, or more direct machine learning scenarios.

Treating cloud as a destination

The cloud conversation has also evolved. It’s no longer about whether to migrate, but how to do it effectively. Yet, I still see businesses committing to hyperscaler contracts without a clear workload strategy. They’re seduced by attractive rates or enterprise commitments, but neglect considerations like vendor lock-in, exit costs, or whether the workload belongs in the cloud in the first place. Yes, ‘lift and shift’ is still a thing, it seems.

The key is workload alignment and recognising opportunities for modernisation. Where, and how, will the workloads perform best in terms of performance, cost, security, and compliance? Hybrid architectures are often the optimal choice, and they require careful planning and a comprehensive understanding of the estate and its dependencies.

Underestimating the skills gap

Both AI and cloud demand new skillsets – from data governance and security, to prompt engineering and threat response. Yet some CIOs still assume that existing teams can reskill overnight. This was never the case, of course, but we seem to forget with every new technology wave. In reality, skills gaps cause disruption and delays, and many AI projects stall or fail because the talent wasn’t in place from the start.

I advise leaders to invest in partnership ecosystems. At Wavenet, we collaborate with specialist partners – including global vendors such as HPE or Microsoft, as well as boutique AI, Power or security experts – to fill these gaps. This approach allows us to give our customers access to the right expertise at the right time, enabling phased resourcing as understanding develops… without needing large-scale hiring upfront.

Data is the starting point

The most successful digital transformations start with a well-defined data strategy. If you can understand how you will use data to achieve specific business goals, such as increasing efficiency, improving decision-making, or gaining a competitive edge, you can then define how and what you will collect, and how you will store and process it. Without a clear data strategy, and effective data processes and controls, then any AI initiatives will likely be hindered. ‘Garbage in, garbage out’ has never been more true.

Stakeholder workshops are often really beneficial in this context. Bringing together leaders from different departments outside of the typical technology setting to discuss what single factor could enhance the business can provide valuable insights and help create a working list of initiatives. This then drives discussion to prioritise the highest-ROI use cases, and ensures the business is on board from the start. Crucially, it moves the conversation from technologies to those all-important business outcomes.

In healthcare, for example, we’ve seen AI used to rapidly triage CT scans. While not replacing human clinicians, this speeds up diagnosis, reducing delays and improving patient outcomes – especially vital given the strain on NHS resources.

In retail, we’ve supported organisations using AI to deliver real-time personalisation. When fed by a well-governed CRM system, AI can drive loyalty and spend, demonstrating how data quality directly impacts success.

Lessons from the frontline

Perhaps the real lesson to be learned in this space is that CIOs should cast off the shackles of classic waterfall project lifecycles and embrace a culture of agile experimentation. Develop quickly, fail fast, learn, cycle round and go again. The tooling and development environments today allow for and, indeed, expect this. If a proof of concept doesn’t work, that’s okay… the biggest risk is not trying at all.

Whatever you do, remember to understand your objectives with real clarity. With that, you’ll unlock genuine value and avoid the pitfalls of being seduced by the technology.

Andy Bevan, head of cloud specialists at Wavenet