AI beyond the hype: How business leaders should think about AI

Interview with Innes McFee, CEO of Oxford Economics

Hosted by Debra D’Agostino, Managing Director of Thought Leadership and Chief Capability Officer, Oxford Economics

Artificial intelligence is rapidly changing how organisations access information, automate work and make decisions. But as businesses move beyond experimenting with chatbots, the question is no longer whether to use AI, but how to apply it effectively.

Debra D’Agostino sat down with Oxford Economics CEO Innes McFee to discuss where AI will create the greatest value, why trusted expertise matters more than ever, and how organisations should think about embedding AI into business decision-making.

AI’s economic impact and the way forward

Debra: 

A couple of months ago, you wrote on LinkedIn about productivity growth estimates from AI and some economists’ more bullish claims. You suggested that a more accurate estimate was around 3% over ten years. Is that still your view? 

Innes: 

Yes. If anything, the evidence over the last couple of months has borne that out.  

AI adoption is certainly increasing, but we still see relatively patchy use across industries. More importantly, most organisations are using AI for a relatively small proportion of tasks, and those tasks tend to be lower value-add. That’s very different from saying AI has fundamentally transformed productivity. 

For AI to generate meaningful economy-wide productivity gains, three things need to happen simultaneously. Adoption needs to be widespread. It needs to be applied to high-value work. And it needs to automate a substantial proportion of those activities. We’re simply not there yet. 

Debra: 

So where do you see the real opportunity? 

Innes: 

I think the biggest change will come when AI moves beyond being a tool that individuals occasionally use and becomes embedded within business processes. 

If AI is simply helping someone write an email or summarise a report, that’s useful, but the impact is limited. Once AI becomes part of an end-to-end workflow, automating multiple connected tasks, it becomes much more scalable. That’s when you start seeing productivity gains that could matter at a macroeconomic level. 

Trusted expertise as the foundation

Debra: 

Many organisations are now turning AI towards their own internal data. But those datasets aren’t always complete, consistent or well structured. How important is the quality of the underlying information? 

Innes: 

It’s absolutely critical. Some business leaders understand that immediately. Others assume AI can simply pull together whatever information is available and produce a reliable answer. 

I always say, if you value your decisions, you need confidence in the information you’re using. If the underlying data isn’t built on consistent definitions and methodologies, how do you know the foundations aren’t made of sand? 

AI is an incredibly powerful tool, and it will become even more powerful. It is very good at assimilating huge amounts of information. But it is not good at original insight, at least not yet. I think it will be a long time, if ever, before it gets there. 

That’s why I think trusted data, rigorous methodology and original analysis will only become even more valuable in an AI world, not less. 

Debra: 

I agree. If AI makes information easier to access, the quality of that information only becomes more important. As organisations begin redesigning workflows around AI, trust needs to become a reliable part of the infrastructure.

The AI questions business leaders should be asking

Debra: 

We’re also hearing from organisations that they’re beginning to redesign how information flows through the business, rather than simply giving employees another chatbot. How do you think business leaders should approach AI? 

Innes: 

I think they should start with the business problem rather than the technology. Too many conversations begin with “How do we use AI?” 

A better question is ” Where could AI genuinely improve a decision, solve a problem or strengthen an existing workflow?” 

Increasingly, our clients are redesigning how information moves through their organisations and rethinking which external intelligence should feed into those decisions. This creates an opportunity to embed trusted economic insight much more deeply into business workflows. 

Capabilities such as the Model Context Protocol, or MCP, are a good example. Rather than asking users to come into our systems, they allow organisations to access our research, data directly through their own AI environments. That’s where I think a great deal of value will be created over the next few years. 

Debra: 

Our research with IBM supports that. Less than a year after ChatGPT changed the conversation, many leading organisations were already well prepared. They did not buy AI and attempt to apply it everywhere. Their approach was use-case first: identifying specific business problems and asking where AI could solve them or help people address them more effectively. 

But that also requires organisations to be realistic about what AI can and cannot do. How should they evaluate those limitations?

Innes: 

They need to be transparent about how the technology is being used, where its limits lie and what has been done to make the outputs reliable. 

Our experience developing AskOE reinforced that. Our economists spent months training and testing the system before it was released. They reviewed outputs, asked the same questions repeatedly, challenged the responses and identified answers that could be misleading. 

We did not simply let the technology run unchecked. We were prepared to give up some functionality to ensure that the information clients received was accurate and grounded in our research. 

That is an important principle for any organisation applying AI. More functionality is not automatically better if users cannot rely on the output. 

Debra: 

With so many potential applications for AI, organisations can easily spread their efforts across too many experiments. What have you learned at Oxford Economics about deciding where to focus? 

Innes: 

At Oxford Economics, I always say there is no point pursuing a thousand initiatives that are all 60% complete if our clients never see the benefit. We would be doing a great deal of work without delivering any real value. 

It is better to choose a small number of important use cases, do them properly and put something in clients’ hands that demonstrably improves their experience or supports better decisions. 

I think the same principle applies to any organisation. The starting point should always be the need you are trying to address. What does the client, customer or user need, and where can AI help you meet that need more effectively? 

That is ultimately how organisations should think about AI. It is not about using the technology everywhere simply because it is available. It is about identifying where it can genuinely improve quality, accessibility or decision-making, applying it honestly and maintaining the trust on which the organisation’s value depends.

---

---