The 23rd EBG | Network summit
In 2008, a category fact base took a team six months to build by hand. With AI in procurement, much of it may now take days. What did those six months produce, besides the fact base?
Six months, by hand. That is how long it took a team to build a category fact base in 2008. The detail comes from the programme text for Hilde Solheim’s keynote on the second day of CPO Outlook 2026.
Her session asks what AI can now do in days, and what still does not stick unless the foundations are repaired. That is her question to answer, on stage.
This piece asks a question beside it, one EBG keeps returning to. When AI takes over work that used to take months, what else disappears with the time it took? And what, if anything, needs to be kept?
What used to take the time
Every procurement team has its own version of the six months. EBG has a record of what some of them sound like.
At a CPO Outlook 2023 workshop, participants were asked what their function spends most of its time on, and what they would remove. Some of the answers are below, verbatim. The full set is published here.
The work behind the fact base
Spend most time on
Would like to remove
EBG | Network, CPO Outlook 2023 workshop with Chr. Hansen and Deloitte, live Slido poll. Answered by 31 and 25 participants respectively. Selection by EBG.
This is close to the work AI is now said to do well. Gathering, cleaning, connecting, summarising. On paper, the wish list from 2023 is being granted.
What the six months also built
Here the question gets less comfortable. A team that spends six months building a fact base also spends six months reading contracts, chasing suppliers and finding the places where the numbers do not add up.
Does that produce something besides the spreadsheet? EBG’s guess is that it often did: people who knew the category well enough to tell when an answer was wrong. EBG cannot prove that, and suspects the answer varies a great deal from team to team.
If the fact base arrives in days, where does the team’s feel for the category come from?
The question is already on the CPO Outlook programme in another form. The Inter IKEA peer lab on Day 1 asks what happens as a process becomes more integrated and simplified: the people running it feel less whether the result is right. EBG wrote about that lab here.
For years, Anna Bjärkerud, founder of EBG | Network, has argued for what she calls Procurement 1.0: knowing your suppliers, and building insight layer upon layer, rather than starting over with every new question or disruption. Seen from there, the slow work was not only a cost. It was also one of the ways the layers were built. Whether AI builds them faster, or skips them, is an open question.
Or can AI carry some of it?
There is evidence pointing the other way too. Erik Brynjolfsson, Danielle Li and Lindsey Raymond studied 5,179 customer support agents who were given an AI assistant. In their study, published in the Quarterly Journal of Economics in 2025, productivity rose by 14% on average and by 34% for novice and low-skilled workers. The most experienced saw little change.
The tool suggested replies in real time, built on what the best agents did. Agents with two months of tenure and the assistant performed about as well as agents with more than six months without it. Some of the gain held even when the tool was briefly down, which the authors read as learning. For the most skilled agents, the assistant may even have lowered quality slightly.
Customer support is not category management. But the finding turns the question around: perhaps AI can carry some of what the slow work used to teach. The authors raise the catch themselves. The tool learns from the top performers, who are not paid for that, and without their input it may get worse at new problems. Whose know-how the AI carries, and who keeps it current, may matter more than the time saved.
Laid on top, or redesigned around?
The programme text describes procurement being handed every wave of technology in the last two decades: new ERPs, automation, and everything that could be layered on top. Economic history has a long view on how such waves pay off.
In a 1990 paper, economist Paul A. David compared the computer with the electric dynamo. In 1899, electric motors made up less than 5% of mechanical drive capacity in US factories. It took roughly two more decades to reach half. The productivity effect in manufacturing did not show before the early 1920s, four decades after the first central power station opened.
Part of the delay, David wrote, was how factories electrified. Many first added electric motors to the old system of shafts and belts, which stayed in place. The larger gains came with individual motors on each machine. Then factories could be redesigned: lighter, single-storey, laid out around the flow of materials.
Two ways to bring in a new technology
After Paul A. David’s account of US factory electrification, 1890s to 1920s. Simplified by EBG.
Laid on top
A new motor drives the old shafts. The layout, and the way work flows, stay as they were.
Redesigned around
Each machine has its own motor. The building, the layout and the flow of work change with it.
Paul A. David (1990), “The Dynamo and the Computer”, American Economic Review 80(2): the “group drive” and “unit drive” systems. Illustration by EBG.
David saw the same pattern in his own time. Old paper-based procedures were being kept alongside the new computer systems, sometimes to the detriment of both. He also warned that information structures, unlike factory buildings, do not wear out. Nobody can count on time alone to force a redesign.
The J-curve, then and now
In a working paper first published in 2018 and revised in 2020, Erik Brynjolfsson, Daniel Rock and Chad Syverson name AI as such a general purpose technology. They argue it requires significant complementary investments: new processes, products, business models and human capital. Much of that is intangible and poorly measured. In their model, productivity growth is underestimated in the early years and shows up later. They cite David’s dynamo directly: it took a generation for factory layouts to be reinvented.
Brynjolfsson has since tested the idea on data. With Kristina McElheran, Mu-Jeung Yang and Zachary Kroff, he studied AI in US manufacturing, using Census Bureau data for 2017 and 2021. In a 2025 working paper, they report J-curve-shaped returns: short-term losses in productivity and profitability before longer-term gains. Plants that had adopted AI by 2017 grew faster in revenue, labour productivity and employment up to 2021, among those that survived.
The losses were concentrated in older businesses, and smaller in young, growth-oriented ones. In the older establishments, roughly a third of the loss came from structured management practices falling away as AI came in, such as monitoring key performance indicators and production targets. The study covers industrial AI in factories, such as machine learning and machine vision, not generative AI in procurement. In EBG’s reading, it still rhymes with David: the cost showed up where existing ways of working had to adjust.
In EBG’s reading, that is one way to read the word “foundations” in the keynote text. Not only data, but the processes and people a technology needs around it. Which of procurement’s past waves were laid on top, and which were redesigned around? Each function will have its own answer.
Where the Nordic rooms say they are
EBG’s own surveys suggest most Nordic functions are somewhere in the middle of that story.
High expectations, unfinished foundations
EBG | Xperience 2025 survey, 86 procurement professionals from 81 companies; the AI barrier question was asked in Copenhagen only (27 participants). EBG | Xperience 2026 survey, 48 senior procurement professionals.
The Xperience 2025 report found most functions past the basics, but struggling with integration. In the Xperience 2026 survey, expectations of time saved by AI were far ahead of today’s reality. Where that saved time tends to go is a question of its own.
One of the questions the keynote puts to the room follows from this. Data is finally flowing, so why is category management back on every agenda?
On stage on Day 2
From shiny object to something that matters — what AI changes in procurement, and what it does not
Hilde Solheim takes 23 years in procurement and condenses them to one page: what changed with each wave of technology, what never did, and what that means for the choices a CPO faces now. Until August she led procurement at Eviny, following 21 years at Equinor. She is now Director at Deloitte, working with procurement and supply chain leaders across the Nordics.
What will be explored, in the programme’s words:
- What are procurement’s non-negotiables when AI promises to do most of the work — and where are the real opportunities?
- AI can now do much of the data tidying that took a team six months in 2008 — so why does it still not stick?
- Data is finally flowing — so why is category management back on every agenda, and what happens to the companies skipping the basics on the way there?
Hilde Solheim | Director | Deloitte AS · previously Head of Procurement, Eviny GroupThe keynote is enabled by Ivalua, which EBG welcomed to CPO Outlook 2026 with the foundations question in mind. Ivalua also hosts a Theme Discussion on Day 1, where participants work through the move from AI pilots to daily operations with peers.
From Pilots to Real Work: Scaling AI into Your Daily Procurement Operations
Theme Discussion, Day 1. Hosted by Ivalua, Henrik Nyberg, Sales Director Northern EuropeFoundation, speed, value — developing procurement tech at enterprise scale
Peer lab, Day 1. Kristina Prilepchanska, Head of Procurement Solutions, Electrolux Group. Read the articleProcess development, digital enablement, the line: who develops, who enables, who implements
Peer lab, Day 1. Mårten Wennersten and Ankit Shah, Inter IKEAWhat to keep, what to stop, what to add — designing UPM’s procurement operating model around business value
Keynote, Day 1. Vesa Kuismanen, UPMWhich road to AI? Who builds it, who owns it — and who answers for it
Panel and room discussion, Day 2, directly after the keynote. Read the article
Three questions to bring to Stockholm
- What was your six months? Name one piece of work that used to take your team months. Who learned what while doing it?
- Laid on top, or redesigned around? For the last system your function brought in, which was it? What would the other option have taken?
- Where will the feel for the category come from? If the fact base arrives in days, how does a new category manager learn to tell when it is wrong?
About Ivalua
Ivalua is a leading provider of cloud-based Spend Management software. Their complete, unified platform empowers businesses to effectively manage all categories of spend and all suppliers, increasing profitability, improving ESG performance, lowering risk and improving employee productivity.
About CPO Outlook 2026
EBG | Network has brought Nordic procurement leaders together since 2010. The 23rd EBG summit, CPO Outlook 2026, runs 14–15 October at Hotel Birger Jarl in Stockholm: keynotes from the stage, theme discussions around the tables, and practitioner-led peer labs. The premise is the same as ever: if you could sit down with peers who share your challenges, what would you want to discuss?
See the full programme · The programme in brief (PDF) · Watch the short film · Register
Sources
- Keynote title, description, questions and speaker line: CPO Outlook 2026 programme — cpooutlook.com/programme-2026
- EBG | Network, CPO Outlook 2023 workshop with Chr. Hansen and Deloitte, live Slido poll — ebgnetwork.com
- David, P. A. (1990), “The Dynamo and the Computer: An Historical Perspective on the Modern Productivity Paradox”, American Economic Review 80(2), 355–361 — jstor.org
- Brynjolfsson, E., Rock, D. and Syverson, C., “The Productivity J-Curve: How Intangibles Complement General Purpose Technologies”, NBER Working Paper 25148, October 2018, revised January 2020 — nber.org. Later published in American Economic Journal: Macroeconomics 13(1), 2021.
- McElheran, K., Yang, M.-J., Kroff, Z. and Brynjolfsson, E. (2025), “The Rise of Industrial AI in America: Microfoundations of the Productivity J-curve(s)”, working paper (US Census Bureau Center for Economic Studies, CES-WP-25-27; SSRN 5036270) — ssrn.com
- Brynjolfsson, E., Li, D. and Raymond, L. (2025), “Generative AI at Work”, Quarterly Journal of Economics 140(2), 889–942; NBER Working Paper 31161 — nber.org
- EBG | Network, EBG | Xperience 2025 Post Survey Report, 86 participants from 81 companies — ebgnetwork.com
- EBG | Network, Xperience 2026 survey, spring 2026, 48 senior procurement professionals.
- Ivalua boilerplate: CPO Outlook 2026 Expert Organisations — cpooutlook.com/expert-organizations
CPO Outlook 2026 · 14–15 October · Hotel Birger Jarl, Stockholm