Two chairs, one process: what the room maps on day two

Procurement is asked to govern AI for the rest of the business and rebuild its own function with the same technology, in the same week, with the same team. One keynote brings the view from outside procurement: how Nordic enterprises actually use AI, what scales and what stalls. One peer lab supplies a method for the function itself. The working session maps where each function stands, and the closing keynote asks what happens when a company stops piloting.

A CPO Outlook 2026 discussion piece · EBG | Network

Most conversations about AI in procurement start in one of two places. Either with the tool, which platform, which feature, which agent. Or with the big picture, how the function will look in five years. Both are legitimate. Neither tells a procurement leader what to do on Monday.

The CPO Outlook 2026 programme takes a different route, and it is worth reading two sessions together to see it. The Day 2 programme opens by naming the requirement: two jobs at once, with no second team to do the second one. Nina Rapp’s keynote then brings what a procurement room rarely gets, an outside view of how Nordic enterprises are actually using AI, from assistants to agents, and why some rollouts scale while others stall. Marjo Jokinen’s Day 1 peer lab describes a way of working through it inside the function: start in the process, not in the tool. The working session that follows the keynote is where the room puts the two together, table by table, function by function, with Tobias Lindberg’s Day 1 lab on who to hire in 2027 as the third stretch of the road. And the day closes with Mauricio Bozzo of Arrive, who has stopped waiting for the ground to firm up and explains what that decision looks like from inside a company.

Two chairs, same week, same team

The programme’s framing of the day is blunt. Procurement is being asked to do two things with AI at the same time. Govern AI for the rest of the business, the licences, the contracts, the agents, the renewals nobody modelled properly. And rebuild its own function around the same technology, with new ways of working, new roles, and a data foundation that has been work-in-progress for years and is suddenly what holds everything else back.

That is not a technology problem. It is a capacity problem with a technology label. The same people who negotiate the enterprise AI agreement are expected to redesign their own category work around what that agreement makes possible. Two chairs, same week, same team.

What the keynote adds is not a prescription for procurement. Nina Rapp does not work in a procurement function and does not pretend to. She works with enterprise AI rollouts across Nordic organisations, and what she brings is the pattern she has seen across many of them: where usage actually takes hold, what the economics look like once it does, and what tends to stall it. For a room that is being asked to sit in both chairs, that outside view is the useful one.

Where the outside view meets the function

The keynote’s premise is that of all the transformation patterns AI produces inside a company, the one now visible across almost all Nordic enterprises is how work itself is being reshaped, from assistants to agents. That is where the economics change. It is also why the outside view matters for both chairs at once: the way the business adopts and governs AI shapes what procurement can do with it, and what procurement learns in its own function is what it brings back to the enterprise conversation. The keynote describes the enterprise pattern; the rest of the day is where the function works out its own. Jatin Singhal’s Day 1 lab, From gate to shaper, takes that enterprise side further: procurement’s place in the corporate AI agenda, and where the function-defining decisions get made.

One step earlier than the tool

Marjo Jokinen’s peer lab on Day 1 starts from an observation that will be familiar to anyone who has sat through a vendor evaluation: most procurement leaders meet AI as a tool question. The lab argues that the more useful question begins one step earlier, in the process itself. Where in strategy, category work, source-to-contract or procure-to-pay does AI change what is actually possible, and what is the gap between where a function is now and where it is heading?

The reason a working example matters is in EBG’s own data. Across EBG’s surveys most respondents rate their digital maturity as “developing”, and few have concrete cases in production. A method that starts from a live case, and from a process the participant already owns, gives the table something real to measure the gap against. A method that starts from a feature list does not.

What the gap looks like in numbers

EBG’s Xperience 2026 survey, run ahead of the spring focus days, gives a rough outline of where Nordic procurement functions stand. In the Stockholm cohort, 78% described their digital maturity as “developing” and none as “leading”, up from 74% “developing” the year before. Asked about urgency to change how the team works, 48% said they were already moving with concrete action, and a further 23% reported high pressure and momentum. When the room was asked who uses AI tools in their daily work, every hand went up.

The hours are the clearest measure of the gap. Across the full Xperience 2026 sample, 29% save four or more hours a week with AI today, and 67% expect to within two years. The savings are real, and the expectation is set well above them.

Put those numbers next to the keynote’s two chairs and the tension is obvious. The hours are expected to arrive. What is not yet decided, in most functions, is which chair they go to.

Where the freed time goes

This is the question EBG keeps returning to, and it is worth being honest that the surveys do not answer it. They measure hours saved and hours expected. They do not measure what those hours became. The Xperience 2026 barriers question offers a clue: 62% of respondents named competing demands as what stops them acting when the opportunity is already clear, ahead of stakeholder alignment, mandate and data. Capacity, in other words, tends to be absorbed before anyone decides where it should go.

The workshop tables at Xperience Stockholm added another clue. Asked to place five procurement tasks on a line from human-led to AI-led, today and three years out, very few dots reached pure AI-led territory even at the three-year horizon. The room’s picture of its own near future is human plus AI, not AI instead of humans. That has consequences for what a process stage looks like once AI is in it, and for who runs that stage.

Which is exactly what the working session on Day 2 is built to draw out.

How day two turns it into a map

Requirement, method, consequence

Notice what the sequence does. The keynote supplies an outside reading of how AI use actually spreads inside an enterprise. The first stop asks each table to apply that reading to itself, in production, not in plan. The second stop takes the Day 1 lab’s process map and asks the harder follow-on question: not where AI changes the work, but what the person running that stage now needs to be able to do. The third stop, held by Tobias Lindberg of Sandvik SRP, moves that to a hiring decision: his Day 1 lab writes the job ad for the 2027 hire, and the working session asks how much of that capability each function already has, and in whom. Requirement, method, consequence, in one working session.

What each participant leaves with is deliberately concrete: their own function two years out, the work, the roles that carry it, one capability to build in someone they already have, and one to go to market for.

Then the day asks what happens if you stop piloting

The sequence does not end at the tables. Day 2 closes with Mauricio Bozzo, Group CIO of Arrive, and the closing keynote is the deliberate counterweight to everything before it. The programme text puts the case plainly: EBG’s own surveys say 74% still struggle with poor data, around 5% have integrated AI into their decisions, and not one organisation called itself digitally leading. High ambition, thin foundation. The reasonable conclusion is to wait until the ground is firmer. He is doing the opposite, and he closes the summit by explaining why, and what the pilot-and-evaluate approach is quietly costing the companies running it.

There is a reason this voice closes a procurement summit. Mauricio spent ten years in procurement before moving into technology leadership, including a role as Global Purchasing Director at Electrolux reporting to the CIO and the CPO at once. He has sat in both of the keynote’s chairs. Read in order, the day runs from an outside view of how AI use spreads, to a method for the function, to a map of where each function stands, to a company that decided not to wait for the map to be finished. Whether that last step is brave or premature is exactly the argument worth having with the people at your table.

The question, then

The programme is careful not to promise that AI will make procurement’s two jobs into one. It will not. What it offers is a sequence that most functions never get: a clear statement of the requirement, an outside view of how AI use actually spreads, a method that starts one step earlier than the tool, and a room of peers that will say, one line at a time, where they actually stand. The gap between “developing” and “leading” is not closed from a stage. It is closed one process stage at a time, by people who know which stage they own.