I am a big believer in procurement getting involved as early as possible in new product development based on all the market knowledge available. The worst-case scenario is product engineering developing a new product, selecting a supplier and then telling procurement to get on with negotiation.
That’s why I was fascinated to get Lora Cecere’s insights into how organizations need to adapt their approach to NPD, involving procurement from the start. We also examined where AI can help in this new procurement reality, and where, as humans, we must “unlearn” the old way of doing things and focus on what AI cannot do.
Lora is the Founder of Supply Chain Insights LLC. She works with supply chain leaders to take teams to higher levels of excellence and is an influencer with more than 340,000 followers on LinkedIn.
She agrees that we should design the supply chain in the stage-gate processes in alignment with what’s going on in the market. Organizations should also be thinking about how to get from quality of design to quality of performance. It’s an area where procurement needs to be very active, yet according to Lora’s research, less than a third of companies have a supplier development arm to help suppliers understand what the design means. The more a company introduces new products, the more it needs to manage these relationships closely, especially as we now, more than in the past, need to consider dimensions such as sustainability and ethics.
If the company strategy is to grow, within the current environment of huge variability, we need to move away from functional metrics on cost to the balanced scorecard at the intersection of growth and asset strategies, because if we don’t have the right assets we can’t grow, and if we are only focused on functional cost, we won’t be able to serve the customer. Growth has the highest correlation to market capitalization in public companies and customer service is a fundamental driver of growth. The ability to manage variability from market to market requires the ability to make trade-offs, particularly in the tactical and operational planning horizon, which most companies are unable to do with current processes and taxonomies.
How well you integrate the suppliers in your processes of course depends on how well you perform as a trading partner you are. Do you give suppliers accurate forecasts? How good is the quality of your design documents? How often do your purchase orders change? On average a P.O. changes 3.8 times and the greater the frequency of change, the more noise is created in the system, which means more issues around matching and synchronization. In this highly variable world, it is more important to focus on synchronization and interoperability than integration alone.
Listen to your suppliers
Organizations must do a better job of listening to suppliers across the full spectrum of their operations. Procurement must take a lead in this, of course, but there must also be much greater engagement across other touchpoints. In 1992 Harvard Business Review published a seminal article, “Staple Yourself to an Order”. Rather than focusing on the already familiar platitudes about delighting the customer, it addressed “the more humble job of making sure customers aren’t excessively annoyed by the company’s order management processes.” Along with Lora, I would urge people to read it but apply the same methodology to stapling yourself to a Purchase Order. Think about how easy it is to do business with your company. How easy is the onboarding process? How easy is it for your suppliers to get paid? If you resolve these and similar issues you will be amazed how much easier it is for you to get the right quality of goods and services delivered when you need them.
In many organizations we are still in thrall to the almighty finance department. Finance became more powerful with the evolution of the global supply chain, but they tend to lack the knowledge needed to exercise this power. If they are involved in procurement, it is in their nature to believe that you can take decisions based on an excel spreadsheet, which means your company is never going to get full value. So, you must educate finance to understand that, while there are suppliers that can be managed based on cost alone, there are other suppliers that are strategic. You can do this by using technologies such as digital twins to show them the impact of the bullwhip, the phenomenon (discussed in my previous article) in supply and demand chains where small fluctuations in consumer demand at the retail level cause increasingly larger fluctuations in orders and inventory as you move up the supply chain — from retailer to distributor, manufacturer, and supplier. For example, if your supplier is on wafer-thin margins, it may be bulk purchasing in order to reduce costs, leaving it with excess stock. This is not conducive to profitable long-term relationships. Demand sensing technologies, collaborative forecasting, and real-time data sharing with the supplier can resolve the issue and rebuild trust.
In an era of variability, the bullwhip effect becomes much more pronounced; during Covid it was at its most brutal. But we have seen other issues amplifying variability with a negative impact, especially in logistics. These could be climate-related, such as barges being unable to move on the Rhine because of low water levels. Or they could be economic related, such as the number of bankruptcies among freight companies (due to overcapacity, falling rates, labor shortages etc.) Or again, geopolitical factors such as the Middle East conflict threatening shipping lanes. If you have adopted a logistics sourcing strategy based on lowest cost, you will in all probability be one of the worst affected by the bullwhip effect when things get tight.
These considerations mean that the procurement function can never be fully replaced by technology, even with the rise of the intelligent agent with TCO comparisons built into the product. We need to build supplier networks and relationships, and design supplier policies that ensure compliance with the ever-expanding regulatory environment, and we need to synchronize more effectively. If we think of traditional procurement in purely transactional terms, such as writing RFQs and purchase orders, we miss the part that artificial intelligence really cannot manage now and perhaps never can, namely, the relationship part. Moreover, procurement skills must constantly adapt not only in line with technological progress, but also with social and economic trends. Thus, as we transition to a more circular economy, manufacturing procurement will increasingly embrace remanufacturing and demanufacturing.
Conclusion: Where should we focus human intelligence?
Too many promising ideas emerging from design networks won’t break through as long as procurement stays locked in the transactional sphere. Let intelligent agents handle the transactions — and the other remarkable things they can do that humans can’t, such as extracting value from oceans of unstructured data.
But let humans take care of the broader issues. That means unlearning the way we’ve always done things and rethinking how we can do them better — as human beings, with agents supporting us as tools, not taking control. Beyond the usual answer, that we should focus on relationship-building, there are other directions where we can apply our intelligence, once we’ve shed the transactional mindset.
We begin by defining the problem, not just solving it. AI is powerful at tackling well-scoped tasks, but we humans must decide which problems are worth solving, and why. Should we optimize for margin, resilience, or customer experience? That choice sets the direction of everything that follows.
Then there’s the task of designing the operating model. Humans will need to (re)design and evolve the rules of engagement for AI agents: how they interact, how they’re governed, when they escalate. In a variable world, this is an ongoing challenge. We can creatively reshape categories, reframe make-versus-buy decisions, or co-innovate with suppliers. These are all areas where AI can assist but not lead. If we don’t step up, AI will simply optimize yesterday’s logic.
We also bring strategic judgment and scenario thinking. AI excels at optimizing known parameters. But we’re better at navigating ambiguity, competing goals, and complex trade-offs. In a volatile geopolitical environment, who knows what risks lie just around the corner? Data won’t tell you — because this isn’t data. It’s strategy.
And then there’s ethics. Agentic AI can optimize for carbon footprint, but deciding how much environmental risk we’re willing to accept is a human responsibility. The same goes for labor standards, supplier diversity, and human rights. These are not matters for algorithms alone.
AI also struggles with the unexpected. Pandemics, wars, sanctions, black swans — when the data fails, it’s up to us to integrate weak signals, challenge assumptions, and override automation.
Finally, there’s alignment. AI agents will handle workflows. But navigating internal politics, cross-functional trade-offs, and external stakeholder pressure require emotional and political intelligence. We’ve already touched on a perfect example: the need to explain the value of supply resilience to finance — especially when it hits the balance sheet.
In the end, it’s not about what AI can do. It’s about what we choose to do with it.
