AI in procurement needs more than tech: It needs human intelligence
As part of our theme this quarter on AI in procurement – why, how and where it can, or cannot, play a crucial role in the source-to-pay (S2P) process – we have consulted both the practitioners in the field and the tech developers that support them. An often under-heard viewpoint in the conversation comes from the firms that train procurement people in how to embrace these technologies and employ them usefully and competently.
We spoke to Sam Pemberton, CEO of Skill Dynamics, which offers digital procurement and supply chain training for corporate teams around the world. We discussed where and why the human–AI skill gap exists in procurement, and the growing need for both human intelligence (HI) and artificial intelligence (AI) to work together.
He puts this into perspective: “Traditionally,” he says, “when we think about skill gaps, we automatically think about human skill gaps, but AI skill gaps exist too. We all know that just because something is generated by AI, that doesn’t make it perfect. In the same way that we humans need training, we have to make sure the AI systems are trained and work well too.”
A recent Skill Dynamics paper focuses on the ability of human intelligence to drive artificial intelligence and interact between the two.
“Many companies we have encountered push full steam ahead into implementing an AI system without really thinking through what needs to be achieved and how. Take strategic supplier relationship management or category management, they both require commercial judgement, which is better done by humans. And while AI is very good at many things, humans still have to be trained to be able to interact with the systems they are provided with.” So …
Human judgment remains critical
“What AI is really good at is pulling together disparate information for the human to consider. That might be parsing big data sets or data utilization, or it might simply be about stocking-keeping levels. But the human needs to be involved more with the subjective: what can be done with the information I’m presented with? Here, the human has the broader experience and understanding of the context than an AI system.”
Sam considers this ubiquitous, across source to pay and the entire supply chain, with the key being to use both human and AI where they work best.
According to the paper, research by Accenture shows that up to 40% of procurement work is currently being augmented by AI, primarily in data interpretation, supplier analysis and workflow automation. And the most mature use cases from the ProcureCon CPO study are in supplier risk monitoring (60%), contract lifecycle management (58%) and invoice automation (57%). Yet 69% of leaders feel only “somewhat confident” in their current strategies and operating models, with just 4% “very confident”. Only 10% of teams feel “highly prepared” to adopt AI and digital tools at scale. So why is that?
Limited training budgets for AI
Despite the rapid rise of AI adoption in business, only 29% of companies said they had a specific AI training budget. “So the skill gap is important to address,” says Sam, “not least because as AI agents continue to do many more things, their output will still need governance and validation from a human.
“Skill gaps have always existed, they exist today and they will exist in future. But this will be especially impacted in the future if people lose historical experience, which is likely as more jobs are handed over to AI. So it’s important to start training now so that human intelligence and artificial intelligence have a better chance of working together, because you have to keep humans in the loop.”
While Sam doesn’t believe AI will replace the source-to-pay professional, he does see that those who embrace AI will replace those that do not.
“During any major inflexion point with technology, those who embrace it do better than those who don’t. History tells us that’s what we will see with AI. Take the IT evolution more generally; the people in the ‘70s and early ‘80s who preferred to keep the typewriter, did in fact adapt to the word processor. But we expected to understand an awful lot more now, be able to drive new systems, make the most of them and create competitive advantage.”
“Today, you only have to look at the sophistication of procurement systems in place in the average procurement setup of a global organization to know that procurement is keeping up. But there is a finite number of people who understand those systems. Being taught how to do something with a tool isn’t the same as being taught how to get the most out of it in terms of the company’s needs. It comes down to having the skill sets. Take strategic supplier management — it’s not about clicking a button, it’s about building and understanding relationships. And because there are many aspects of that, the training must be very nuanced. Managers want that understanding to be more evenly dispersed among their team. Given that 69% of respondents said there’s an over-reliance on on-the-job training, it shows that managers don’t generally have the time to focus on individual training needs. But it’s also important not to just train for training’s sake.”
Employ AI for the right reasons
Training is a way to future-proof skills and knowledge and to work with AI to its best advantage. “Employing agents as a way to reduce headcount is not the right move,” says Sam. “It might bring short-term gain, but it’s not sustainable. The day won’t come when we turn the lights out, leave the office and let the computer run everything. It’s not realistic. What happens when we remove procurement apprenticeships but expect to have CPOs with judgment?
“And there remains the question of company culture. For most companies, there will always be the need to work with their teams, understand them and what they are trying to achieve, and then work out how to do that in the widest sense. Companies that embrace training understand the need to have skills spread across the organization; visionaries will not be the ones getting rid of their people. There might be a shift in the focus of humans, and in some industries, a change in the number of humans, but the real focus will be on training people to understand the systems and ensure they are operating as they should.
“We will always need people who can think things through and see where they are going right, or wrong. The point is, AI can be extremely useful, but it has to be used in the right way alongside human judgment. Otherwise, you risk having to hire all that talent back again, and probably at a higher cost and with less of the experience you’ve just lost.”
Plugging the skills gap
Part of the gap problem, according to the paper, is that only 10% of teams feel confident they are prepared to scale AI effectively.
“If a massive 90% don’t feel they have the confidence and readiness to adopt, that should be concerning for a lot of business leaders,” says Sam. “So training is crucial. We conduct a unique review for every learner, whether they are in supply chain, procurement or trade compliance.
“We are able to gauge what good should look like in their discipline by factoring in many other capabilities across the hundreds of thousands of learners globally that we have on our system. We discern what skills they should have and where the gaps are. Through a scoring system, we can see where the learner has a high or low understanding of certain areas. For example, if the learner scores low on legal understanding, the system will drive legal training with highly specialized, highly impactful training that brings that person up to the level they need to be at. That might be benchmarked against other employees or against job requirements. Increasingly, the gap we are working mostly to plug is in AI understanding. So we look at how AI specifically relates to their job rather than teaching about AI as a competency. We model AI within each area, looking at where it helps or hinders, what the learner needs to understand and the specifics that are important for them to carry out their job (see Figure 1).
Fig. 1Identifying the skill gaps – example
Source: Skill Dynamics
The courses are created by in-house subject-matter experts who are often ex-practitioners themselves. As such, they have a firm understanding of how practitioners learn differently, have different capacities to absorb leaning, and have different timing requirements.
“AI and its possibilities are prolific in this new era,” concludes Sam, “but it doesn’t mean we can forget everything we’ve learned as humans so far.”
For more on AI in Procurement visit our dedicated collection of research, analysis and case studies.
For more on procurement training options, consult The Modern Procurement Training landscape.
