AI in practice

Don't let the tail wag the dog

Few boardrooms have escaped the instruction. At some point in the last two years, someone senior has said: "We should be doing something in AI."

It is an understandable request. It is also a classic case of the tail wagging the dog.

Left unchecked, it sends a group of tech-literate (and some less tech-literate) people off to build something that may be very impressive, but which is largely disconnected from what the business actually needs to secure competitive, sustainable growth.

There are some parallels with the early days of internet and e-commerce adoption, when many organisations spent considerable time wondering if the new technology really was a shift in paradigm, and whether (and how) to jump in. Here the parallel ends.

Organisations know AI is coming, and they know there are huge benefits to reap. Ambition is not the problem. Execution is. And execution stalls so often because organisations start in the wrong place.

There is always a box

In many ways, AI presents the same challenge as innovation. Saying “we want to innovate” is not a useful starting point. Innovation should always be in service of growth, top or bottom line. Define how much growth you need, and how fast, and you have something against which any idea can be measured.

A blank sheet of paper and encouragement to think 'outside the box' is nearly always a waste of everyone's time. There is always a box. The challenge is to innovate within it.

AI is exactly the same. Ask "what do you want AI to do?" and the answer is usually "well, what can it do?" – and the conversation goes round in circles. The better question is "what do you want your business to do, or to be?" Where would you need to be to make the biggest competitive or operational difference? Work back from there.

That is a strategic conversation, and it needs a clear 'why'. It also needs to stay realistic. Customers have to come with you – leaping to a future that neither you nor they can relate to is not progress.

A layer, not a silo

The second parallel with innovation is that AI cannot be bolted on as another silo, or an extension of the IT and digital teams' box. To realise its potential, it needs to be understood as a layer between the processes and the people, cutting across the whole organisation.

That means AI is not IT's domain alone. Delivering a strategically significant outcome is never the job of a single function. Businesses are complex, organic systems, and working in the space between what is desirable and what is possible needs people from different functions in the same room.

It also changes the role of technology leadership. The CTO's job becomes less about managing a tech stack and more about designing business processes. Whether to build a system or buy into one, whether to develop specialist 'vertical' agents grounded in the organisation's own context or more generalist 'horizontal' ones – these are strategic decisions and as such, should not be the domain of any individual group.

Redesign the work, not just the tools

Once the outcome is clear, the real work begins: redesigning the processes underneath it. How information is gathered, analysed and moved. Where decisions are made. Where approvals sit. What hand-offs exist. All of it can be pulled apart and reassessed, with the business outcome as the guiding light and AI as a key component enabling competitive advantage.

Data is often cited as the reason to hold back. Quality matters, but not all of your data needs to be 'gold standard'. You need enough good data for a model to learn what good looks like – it can then help sort out the rest. What you cannot do without is confident, competent people, and the governance to oversee outputs before anything is put in front of customers or allowed to make decisions. Regulation adds a further key filter, particularly where data crosses US and EU jurisdictions.

No excuse to wait

At the time of writing, new AI models are being released roughly every ten days, and it is tempting to hold out for the next, better one. But the platforms are already capable of doing what most organisations need them to do. The constraint is not the technology – it is inertia.

There is a distinction worth drawing, though. A strategic programme to redesign how the business works is a long way from giving people permission to 'play' with an AI tool on their desktop – which, without care, can be risky. The pragmatic answer is to allow both. If something makes someone's working life easier, let them experiment. If they want to use it at scale, it must go through a diligent and thorough approval loop.

Putting leaders into AI

Perhaps the biggest challenge is human. Attitudes to AI span the full spectrum, from fear to excitement. Some want to race ahead; others want to wait and see. That range is entirely natural and reflects the richness of the human condition. But leaders need to understand where they sit on this spectrum themselves, because their mindset cascades through the organisation.

There are of course uncertainties. Beyond the somewhat metaphysical fears of impending annihilation, a more prosaic sensitivity could be that of 'hallucinations' – the risk of AI getting it wrong. To overly focus on them however perhaps distracts from the fact that people make mistakes too, and an agent, once corrected, will never make the same mistake twice. Are we holding technology to a higher standard than we hold ourselves? And if so, is that about accuracy, or about trust?

AI also unsettles many of the instincts leaders have developed to steer decision making: the avoidance of failure, a drive for certainty, a preference for binary choices, a need to measure ROI. All are harder to satisfy in an AI world. The shift required is less about putting AI into leaders and more about putting leaders into AI – getting close enough to learn, understand and form their own judgement.

The pulley, not the purpose

Steven Kotler describes AI as a way to amplify our intellect, in much the same way that the pulley amplified our strength. It is a useful frame. A pulley is only valuable when you know what you need to lift, and where you need to put it.

The same is true of AI. Start with the outcome. Bring the right people together. Redesign the work. And recognise the human dynamics at play – not least your own. Do that, and AI stops being the tail wagging the dog and becomes what it should be: a powerful means to a clearly defined end.

Tomorrow is coming. The question is whether you shape it, or wait for it to shape you.

This article was created following an April Forum on AI in practice. The April Forum is a regular roundtable event for senior leaders to discuss challenging topics. Visit our Forum webpage for more information.