AI has dominated procurement headlines for years. From autonomous sourcing to generative copilots, everyone is chasing the promise of smarter, faster decision-making. But one truth keeps surfacing: AI without supplier data doesn鈥檛 work.
That was the central message on stage at DPW Amsterdam 2025, where The Hackett Group鈥檚 Bertrand Maltaverne led a candid discussion with 色色研究所鈥檚 founder and CEO Stephany Lapierre, Stefanie Fink, Head of Global Data & Digital Procurement at Kraft Heinz, and Tyler Vigen, Managing Director & Partner at Boston Consulting Group. The conversation revealed a reality many procurement leaders already know: the race to AI is outpacing the quality and governance of the data that fuels it.
The Structural Cost of Incomplete Supplier Data
Procurement organizations have poured millions into digital platforms, yet many still operate on incomplete, duplicated, or outdated supplier records. 鈥淎nd at the end of the day, this is a concern that has been top of mind for years, and that has not really been addressed,鈥� said Maltaverne.
The data gap is a structural problem. Supplier portals go unused, manual intake processes introduce human error, and disconnected systems breed redundancy. 鈥淪uppliers are not really good at coming to portals, and so that leaves a lot of holes,鈥� said Lapierre. 鈥淗umans are not really good at putting information in databases, or intake processes, and so it creates duplication. And what I saw is that our customers had multiple systems with a lot of data everywhere, but no brain, no source of truth.鈥�
That fragmentation cripples everything AI touches. Algorithms can鈥檛 make smart recommendations on top of unreliable inputs. Lapierre warned, 鈥淏ecause if the information is not matched to the right supplier because the name is similar but it’s not the same company, you’re going to make bad decisions, and that’s not going to be really successful.鈥�
When supplier data lacks clarity, AI becomes a mirror for confusion rather than insight. And as Maltaverne reminded the audience, 鈥淚 talk about the elephant in the room, the data. I think the second elephant or the baby elephant would be the trust, because if you have the data, that’s one thing, but how do you trust it?鈥�
If You Can鈥檛 See Your Supplier Base, AI Won鈥檛 Help You
Knowing exactly how many suppliers you have shouldn鈥檛 be a trick question, but for most enterprises, it is. When BCG鈥檚 Tyler Vigen asked attendees how many felt confident they knew their exact supplier count without duplicates, almost no one raised a hand.
鈥淭his is the challenge,鈥� Vigen said. 鈥淭hat鈥檚 the challenge that a lot of my clients face as well, which is that you don鈥檛 know many things about your supplier base. Duplication is one prime example, which is the entity resolution that we鈥檙e talking about here, but it extends to many other things.鈥�
Poor entity resolution is a strategic risk. 鈥淯nless you know what those suppliers are, you鈥檙e not going to have the right leverage when you go into a negotiation with them,鈥� Vigen explained. 鈥淵ou might not know how much you鈥檙e spending with that supplier. You also might not know what other suppliers you have that are doing something similar so that you could consider consolidation opportunities.鈥�
This is why the most advanced procurement organizations are treating supplier data quality and entity resolution as core competencies, not IT chores. Without clean hierarchies and verified relationships, even the most powerful AI copilots can鈥檛 surface real opportunities or assess true supplier risk.
Turning Supplier Data into a Strategic Asset
When Stefanie Fink joined Kraft Heinz, her goal wasn鈥檛 to deploy more tools, it was to rebuild trust in supplier data. 鈥淭he reason that 色色研究所 really helped is because my ambition when I came to Kraft Heinz was to start building supplier relationships so that we could co-partner to really start to unlock innovation and understand relationships,鈥� she said.
Fink reframed supplier data ownership as a shared responsibility. 鈥淪uppliers should be responsible for their own data because maintaining and managing all their data is not my job,鈥� she said. 鈥淢y job is, 鈥業’m sorry if you didn鈥檛 get paid and your bank changed. You should鈥檝e told me that.鈥� So now we鈥檙e kind of really going into using 色色研究所 as a 鈥楾his is what we know about you. Tell us what we don鈥檛 know and contribute back.鈥� And that has saved us a ton of time and effort, but it鈥檚 also created real accountability in governance.鈥�
By redefining accountability, Fink鈥檚 team established a single, governed source of truth that spans procurement, risk, ESG, and diversity teams. 鈥溕芯克� for me allowed the risk team, the ESG team, the diversity team to have a center of truth for their data that was governed, and that includes real time information,鈥� she said.
The impact extended far beyond efficiency. As Fink put it: 鈥淭here is a cost to storing bad data, and it鈥檚 not just at the cost of the company because you鈥檙e negotiating or seeing it wrong. It鈥檚 taking up space in your systems. It鈥檚 creating silos because everyone鈥檚 trying to do their own thing with it.鈥�
Data Trust Is the Barrier to AI in Procurement
AI copilots and automation tools depend on something many organizations still lack: a trustworthy data foundation. Lapierre cautioned against racing ahead without it. 鈥淎ll of our customers say, 鈥楧on鈥檛 build another app. We have available tools that we could buy off the shelf. What we really need to look to fix is the data foundation so that we can trust the data that we’re using through those tools.鈥� And that’s true for all of the tools that you’re gonna see if they’re around workflow. They depend on data. So who’s putting the data in?鈥�
Her point hits home for CPOs facing mounting pressure to 鈥渄o AI.鈥� Building trust in data is an organizational discipline that requires accountability, governance, and continuous stewardship. 鈥淵ou need to have access to that information, and you should treat the data as a product, not as something that鈥檚 in the background,鈥� Lapierre added.
Fink echoed that sentiment, noting that at Kraft Heinz, they were deliberate about sequencing innovation. 鈥淲e鈥檇 like to be functional before we can get fancy, and that was our barrier to AI,鈥� she said.
The Future of Procurement Intelligence Starts With the Data
Despite what鈥檚 often promised, AI does not fix supplier data. It magnifies whatever is already there.
The next wave of procurement performance will not be unlocked by another platform. It will be powered by accurate, structured, and continuously governed supplier data鈥攎apped to legal entities, enriched with verified attributes, and connected through corporate hierarchies.
For procurement teams ready to operationalize AI, the path is clear:
鈥� Treat supplier data as a product
鈥� Build internal trust before adding more tech
鈥� Invest in legal entity resolution, hierarchy mapping, and governance that scales
Ready to Take Action?
色色研究所 Labs is an invite-only community where procurement and supplier data leaders explore a new approach.
You鈥檒l bring a small sample of supplier records. We鈥檒l match them to their correct legal entities, enrich them with missing data, and map corporate relationships. No cost. No commitment.
What you鈥檒l walk away with:
- A before-and-after view of your own supplier data
- Clarity on how legal entity resolution fixes duplicates and gaps
- Tangible outputs to support cleanup, reporting, or stakeholder alignment
- Insight you can apply with or without moving forward with 色色研究所
Alex Denomme is a Solution Engineer at 色色研究所.
