From AI potential to real business value

I don’t think most companies have an AI ideas problem anymore.

Spend a few hours with almost any management team and it is surprisingly easy to produce a long list of things AI might improve. Customer service could become faster, sales teams could get better support, knowledge could be easier to access, reports could be produced automatically, decisions could be informed by more data and repetitive tasks could disappear altogether. As AI agents become more capable, the list expands even further because we are no longer only talking about helping people perform individual tasks. We are beginning to talk about AI performing parts of workflows itself.

There is enormous potential in that. But there is also a danger that we confuse the number of things AI can do with the amount of value AI can create.

I think we need to turn the question around.

Instead of starting with “Where can we use AI?”, I would rather start with “How does this business create value, and where could AI make that value creation better?”

That sounds like a small distinction. In practice, I think it leads to a very different AI agenda.

It moves the conversation away from technology looking for applications and towards the business itself: what it is trying to achieve, how it creates value for customers, how that value moves through the organisation and where there are opportunities to make it better, more efficient or fundamentally different.

That is increasingly where I believe the real AI opportunity sits.

Follow the value

Most businesses create value through a chain of interconnected activities. The exact value chain obviously differs from company to company, but somewhere along that chain customers are understood, propositions are developed, products or services are created, demand is generated, sales happen, things are delivered, customers are supported and knowledge is accumulated and reused.

Within each part of that value chain sit processes, decisions, systems, information and people. Some work extremely well. Others contain friction that has gradually become accepted as simply the way the company operates.

When we look at AI through that lens, the conversation becomes much richer than automation.

There will absolutely be processes where AI can remove manual work, shorten cycle times or lower the cost of delivering something. Those opportunities matter. But efficiency is only one form of value creation.

Take customer service as a simple example. We could look at AI purely as a way of reducing handling time or automating interactions, and there may be a perfectly good business case for doing exactly that. But the same technology might also give a service employee much better access to the company’s knowledge, improve the quality and consistency of answers, anticipate what a customer is likely to need next, personalise the interaction or identify signals that suggest a customer is considering leaving.

The process hasn’t simply become cheaper. It may have become better.

And that distinction matters.

If we look across the value chain with only cost reduction in mind, we will find one set of AI opportunities. If we ask how AI could improve the way value is created, we may find a much larger one.

That is why I increasingly think about AI potential in three simple ways.

Better means improving the value already being created: a better customer experience, higher quality, better decisions, greater precision, stronger personalisation, faster innovation or giving employees better tools with which to do their work.

More efficient means improving how that value is produced: reducing cost, eliminating unnecessary work, increasing capacity, simplifying processes, accelerating workflows or automating activities where human effort isn’t adding meaningful value.

And new is perhaps the most interesting category. AI may make it possible to create a proposition, service, experience, capability or even a business model that previously wasn’t practical at all.

These categories overlap, of course. A better process may also be cheaper. A new capability may create revenue while simultaneously increasing productivity. That isn’t a problem. The purpose of the model isn’t to classify every idea perfectly. It is to broaden the conversation from “How much can we save?” to “How could this business create more value because AI now exists?”

That is a much more interesting question.

The people who know where the value is

There is another implication of looking at AI through the value chain.

You cannot understand the potential by looking at the technology alone.

AI engineers can bring a deep understanding of what models, agents and other technologies are capable of. They can see technical possibilities that most people inside the business will never identify on their own. But they don’t automatically know why a particular customer conversation matters, why a seemingly simple process contains fifteen exceptions or why an experienced employee makes a decision differently depending on circumstances that aren’t documented anywhere.

That knowledge lives in the business.

It lives with the people who work with the customers, products, systems and processes every day — the company’s subject matter experts.

I think they are sometimes underestimated in conversations about AI transformation. We tend to focus on AI expertise because the technology is new, but domain expertise becomes more important as AI moves deeper into the business. If an AI agent is going to participate in a real workflow, somebody needs to understand what good actually looks like in that workflow, what knowledge the agent requires, which decisions can be automated, which exceptions matter and where human judgement should remain.

The subject matter expert understands how value is created today. The AI engineer understands what technology can make possible. Neither perspective is enough on its own.

The work becomes powerful when they meet around a clearly defined business objective.

I like the multiplication sign here rather than three separate boxes because the point is the interaction. Great AI capability applied to an unimportant business problem creates limited value. Deep domain knowledge without understanding what technology can now make possible may miss entirely new ways of working. And both can produce a lot of activity without a clear business ambition telling us why any of it matters.

This isn’t only an intuitive argument. McKinsey’s research into enterprise AI adoption has found that redesigning workflows is particularly associated with bottom-line impact from generative AI, suggesting that value comes not simply from deploying tools but from changing how work is actually performed. BCG similarly argues for cross-functional AI teams that combine technology capabilities with process ownership and business accountability rather than treating AI as a standalone technology programme.

For me, though, the practical implication is straightforward: if you want to understand where AI can create value, get close to the people who understand how value is created.

From a value chain to an AI portfolio

Once you begin working through a value chain with subject matter experts and AI specialists, ideas tend to appear quickly.

That is useful, but a long list of use cases is not an AI strategy.

The next job is to make sense of what has been discovered.

Several apparently separate ideas may actually relate to the same part of the value chain. Five opportunities identified within customer service, for example, may make much more sense when viewed as one redesign of the service experience and the workflow behind it. Ideas found in different functions may depend on the same knowledge or data. One apparently attractive initiative may turn out to be dependent on another capability that doesn’t yet exist.

This is where I think it becomes useful to stop talking about individual use cases for a moment and look for value pools.

Where could AI materially strengthen value creation? Where are several opportunities pointing towards the same underlying change? Which parts of the value chain matter most to the company’s strategy? And where could AI alter not just an individual task, but the performance of a larger part of the business?

This also helps avoid one of the traps I see in AI programmes: a very large portfolio of pilots, each of which makes sense individually but which collectively consumes enormous amounts of organisational attention without changing very much.

BCG’s research has similarly suggested that companies seeing greater AI impact tend to concentrate resources on a smaller number of high-value initiatives rather than spreading investment thinly across many disconnected projects.

The goal, then, isn’t to leave discovery with the longest possible list.

It is to leave with a much better understanding of where AI could genuinely change the performance of the business.

Understanding what an opportunity is really worth

Finding an attractive opportunity is still only the beginning. Before prioritising it, we need to understand how the value would actually be created.

This is where I think AI business cases often become too narrow. If we identify that a process currently consumes 20,000 employee hours and believe AI could reduce that by 30%, it is tempting to multiply the hours by a salary cost and declare the result to be the value of the initiative.

But time saved isn’t automatically value created.

If those hours allow the business to handle more customers without adding people, there may be a genuine capacity benefit. If they enable employees to spend more time on activities that generate revenue, the economic impact could be greater. If nobody changes what they do with the released capacity, the actual financial value may be considerably smaller than the spreadsheet suggests.

The same applies on the upside. A better customer experience may increase retention. Better recommendations may increase conversion or average order value. Better access to knowledge may improve quality and reduce errors. Faster product development may bring revenue forward. An AI-enabled proposition may create a completely new revenue stream.

The job of the business case is to make the mechanism of value creation explicit.

At the same time, we need to understand what the opportunity depends on. The necessary data may not exist in usable form. Systems may need to be connected. Knowledge may sit primarily in people’s heads. Processes may have to change. New governance may be required. Employees may need different skills, and roles may change as the division of work between people and AI changes.

People impact therefore belongs inside the business case, not in a change-management plan that appears after the technology has already been designed.

An AI initiative can remove frustrating work, give people better information and increase their ability to make good decisions. It can also fundamentally alter roles or reduce the need for certain activities. Both possibilities need to be understood honestly if we want a credible view of value.

So rather than reducing the AI business case to a single ROI calculation, I would look at it as a connected picture.

This isn’t intended to turn every AI idea into months of analysis. Quite the opposite. For many opportunities, the first business case can be deliberately lightweight. What matters is that we can explain how the idea is supposed to create value and what has to be true for that value to materialise.

If we can’t do that, I don’t think we are ready to prioritise it.

Prioritisation is where strategy begins

Eventually, somebody has to choose.

That sounds obvious, but it is where AI programmes can become surprisingly uncomfortable. When dozens of possibilities have been identified and different parts of the organisation are enthusiastic about their own ideas, prioritisation inevitably means saying not now to some very reasonable opportunities.

That is precisely why the work before prioritisation matters.

We can compare opportunities not simply by how exciting the technology is, but by the value they could create, the confidence we have in that value, the strategic importance of the part of the value chain they affect, their dependencies, the investment required and the time it may take to produce meaningful results.

Some opportunities will create relatively modest value but can be implemented quickly and build organisational capability. Others may be difficult, expensive and dependent on significant change, but have the potential to transform a strategically important part of the value chain. Those are different bets and should be treated as such.

I would much rather see a company make a deliberate commitment to a small portfolio of meaningful AI initiatives than launch dozens of pilots because nobody wanted to choose.

That doesn’t mean we need certainty before acting. We won’t have it. It means understanding enough to know which uncertainties are worth resolving.

And that is where experiments, prototypes and proofs of value become useful. We can test the assumptions behind an initiative before committing to the full transformation, then update the business case as the evidence becomes stronger.

The process therefore isn’t a funnel that produces a fixed roadmap. It is a way of progressively improving the quality of the decisions we make.

And as the work progresses, the portfolio should narrow. We may begin with many possibilities across the value chain, group them into a smaller number of meaningful value pools and eventually commit to a few initiatives for which there is a clear reason to believe the impact justifies the investment.

That is a very different outcome from a spreadsheet containing 150 AI use cases.

AI should make the business better

The more I think about AI transformation, the less I see it as a separate technology agenda.

AI is a new and extraordinarily powerful capability, but ultimately it enters a business that already creates value through customers, products, people, processes, knowledge, technology and economics. The interesting question is what happens when that capability begins to change the connections between them.

Some processes should become dramatically more efficient. We shouldn’t apologise for that. Removing unnecessary work and reducing cost are legitimate forms of value creation.

But efficiency on its own is a limited ambition.

AI should also help companies make better decisions, create better customer experiences, improve quality, innovate faster, give people stronger capabilities and develop products and services that previously weren’t possible.

In some parts of the value chain, the biggest opportunity may indeed be doing the same thing at half the cost.

In others, it may be doing the same thing twice as well.

And in the most interesting cases, it may be doing something the business simply couldn’t do before.

That is why I don’t think the most important question for leaders is:

Where can we use AI?

I think it is:

Where in our value creation could AI make us meaningfully better?

Once that is clear, the technology conversation becomes much more useful. Subject matter experts can bring the reality of the business. AI engineers can bring an understanding of what has become possible. Business Design can connect those perspectives, make the value explicit and help turn possibilities into decisions.

That is how we move from AI potential to real business value.