I’ve been thinking about AI ever since I learned about neural networks. It was nearly 30 years ago. I spent some time working with a guy who was using them to develop a better way of predicting stock markets using AI. That was in 1994. It wasn’t called AI then – it was just called ‘computers’. Ever since then I’ve seen great claims made for artificial intelligence. It’s an idea that has been born more times than 3D cinema and the internet fridge and, as you can tell, I’m somewhat sceptical.

This happened with the cloud. As a concept it was compelling – software as a service was the natural next big thing after the internet matured and as the model became more credible so every solution provider claimed to have a cloud model. It muddied the waters and made it difficult to discern what cloud actually meant. But 2 decades on we are in no doubt what cloud is, that it is a real thing and it’s important. But can we say the same for AI?

On one level, this is a dumb question. Of course AI exists, it is real and it is important. There are numerous examples in the real word – Siri and Alexa and shopping recommendations spring to mind as obvious applications that are already embedded into daily life. But what about AI in business and in particular in the sphere of procurement and purchase to pay? There is no shortage claims for the application of AI but there always is when there’s a new buzzword that will sell strategy consulting. How much of it is real and how much is hype?

When sourcing commodities globally, learning from patterns in big data to make recommendations that would otherwise be impossibly difficult is a great example of AI that is being applied today. Extending the fraud detection techniques that the banks use can be applied effectively to capture potential P2P fraud. And there are a number of good applications for RPA in enhancing back-office processes. But is RPA even AI? Robotic Process Automation is in my opinion borderline AI but I wouldn’t want to get too drawn into an argument about it and I think this is part of the problem – agreeing, without getting into an academic debate, what we understand AI to be. It’s like asking “What is Art?” I can hear myself saying “I don’t understand AI but I know what I like.”

There’s a Venn diagram with AI experts in one set and procurement, S2P and P2P experts in another.  The intersection is sparsely populated. GEP recently published a paper that gives some very good examples of the application of AI to Supply Chain and Procurement including a breakdown of how it can be applied in S2P. While I wouldn’t have any criticism for the paper itself – it’s thoughtful, thorough and expert and I would strongly recommend it, most of the examples quoted of where AI can apply to S2P rely on there being no S2P systems or processes already in place.

To quote the paper: “Procurement managers are aware that there are cost-saving and improvement opportunities across their supply chain. However, without the full picture, they’re unable to fully comprehend what spend is avoidable, and what costs are actually enhancing the business’s success.

“A common challenge with consolidating spend data is that disparate tools used across the businesses will typically apply different categorizations to the same types of spend”

I couldn’t put it better myself. But this is what a S2P system addresses. Solutions like Coupa, Ariba Jaggaer, Basware and the like address these business issues. Of course, AI could be applied to take messy data from disparate systems and glean valuable insights – or you could tidy your data and perform sourcing and procurement on a single platform.

It’s a bit like saying that you don’t need a filing system when you have AI. It’s true, but the converse is also true. Why invest in AI when you can spend much less in a more established filing system.

These use cases for AI make sense in an environment where spend is not controlled, where S2P does not prevail. But if an organisation hasn’t got an S2P system, it doesn’t need AI, it needs a S2P system. In these examples, AI is a solution looking for a problem.

Although I am sceptical about the application of AI to procurement and especially to P2P and S2P, I’m not without some enthusiasm for it. There are one or two relatively obscure use cases that I can see having legs but most use case I fear, have very little merit. There are those that think that AI could enhance processing of paper invoices and extraction of information from contract documents but paper invoices are being rapidly superseded by electron documents (which will themselves be superseded by blockchain at some point) and extraction of contract documents is, well, nothing new.

And at the end of the day artificial intelligence is another word for bigger and better computers. All that AI will do is to provide better information and insights with which to make business decisions. This is a good thing of course but the thing the technology cannot do is to actually make the decisions. AI will make it easier and quicker to make choices – but you will still have to make the choices. A computer can’t tell you how to run a business any more than a telephone can tell you what to say.

When you look at the most impressive examples of AI like deepfake and face recognition, they get very close to passing for the real thing. My phone recognizes my face as effectively as a human and a deepfake video presents to me what appears to be Tom Cruise clowning about, and even though I know it’s fake, I can’t see the joins. These examples, to a layman like me, pass, or at least come very close to passing the Turing test, the test of a machine’s ability to exhibit intelligent behaviour equivalent to, or indistinguishable from, that of a human. Even the best examples of AI applied to supply chain and procurement don’t come close to that. So for me, for the time being at least, while I remain a technology enthusiast, the opportunities for AI in Supply Chain and procurement remain pipedreams.