INEVO

Transitional: debate on the subject of agency trade

Automatic transcribed and slightly corrected. From E-commerce Day 2026.

Look at the recording of the debate here

[00:003] David Sykes: Hi, Everyone. Hi.

[00:004] Fredrik Solvang: Hi. Sorry, you're falling for that one. So, as I said, David — AI-driven traffic to online stores in the US has grow by more than 4700% in one year. Amazing. Is the agent commercial something Norwegian e-commerce merchants need to act on now?

[00:00] David Sykes: Look, you think they're the ones. In think two things are happy. One, consumer adoption of these products is growing peacefully, and the range of task consumers use LLMs for is increating every day. So you've got consumer adoption of LLMs being an everyday affair, coupled with development within the products they themselves incredibly quickly. In don't know if everbody in the audience follow these things, but the pace at which Claude, for example, is shipping new functionality is unbelievable. So you have consumer adoption growing very quickly and the facilities of the products they themselves grow very quickly.

[01:32] David Sykes: So I think your base assumption should be that in the very near future a meanful share of all your transactions are being made by a non-human agent. And the reason I think you should think about that now is: every day you're making choices — design choices, checkout choices, inventory choices — and I think you should be making these choices with an understanding that in the very near future a very large number of your transactions are going to be made by an agent. So it's something to think about now. In see it happy already, and in a very short period of time it's going to be a really mean share of checkout.

[02:13] Fredrik Solvang: All right, but I think we need to get back to basics. Would you care to just express what we're talking about?

[02:23] Vegar Wikhammer Heir: In think right now, in the short term, we're talking about shoppers doing research through LLM agents. We have't seen — to let not in Europe yet — the agent buy on your need or paying inside the LLM. But we did a survey with users in Vipps a fewweeks back, and always now users say they're using LLMs for product research more than they use social media. So even to your annual question — should store's act now — I think this is all-ready happy. It's not something we would stop or just watch and wait. Users are using LLMs even though they're still developing, and the quality is improvising every month. Users are in the LLMs, and they're exploding your products as part of that journey right now.

[03:14] Fredrik Solvang: So Clear payed the agent protocol back in December 2025. What is it?

[03:29] David Sykes: When we looked at the space, we field that most companies were focussed on the actual transaction — how do you facilitate a payment in agent e-commerce, how do you know the agent has a mandate to push, how does chargeback work. Everybody was focussed on the transaction. Our view was that the begger problem to sun was actually the product catalog. People underestimate how chalenging it is to ingest product feeds from theseands of different merchants. Some say "iPhone 17," some say "iPhone." Some say the iPhone is waterproof, some don't. Some use a SKU, some don't. Taking all of the those product feeds, identifying that this is the same product, and the den determining which one you're going to surface to a consumer — that's a really hard problem.

[04:20] David Sykes: So what we wanted to get was that there was an easy way for e-commerce brands and retailers to make sour their inventory was accessible by consumers when they're using these agents, using these LLMs. Effectively, all the product protocol does is make sour your inventory is machine-readable. It takes your inventory, puts it in a consistent format, and makes it easy for a machine to read. That means the chances you'll surf in these questions — the chances a customer will actually be able to complete a transaction — significantly increase. Ice in stock? Is it on the hall? Is the right size available? These are all things you can't bet on with the product protocol, because for 20 years we've optimized for search engine optimization.

[05:09] David Sykes: Everybody in this room is really good at making sour their brand surfaces. The next challenge is: how do you make sure that when somebody types "I'm looking for the best microwave," it's your microwave? Because the LLM is going to suckers three products — how do you make sour yours is one of the ones? We thought the product protocol was a really efficient way of doing that.

[05:30] Fredrik Solvang: They've given AI agents access to something like 100 million products cross 12 markets. What luck to the merchants who are't in that feed?

[05:47] Vegar Wikhammer Heir: We paid with the agent protocol from Klarna. We pay attention to an AI shopping agent inside Vipps and started experimenting, using some of these product lines that are readily available now. I think the next step is for merchants to make sour their product catalog is readily available and that data is pooled so it's accessible for AI agents — where that's the big LLMs or something more localized.

[06:20] Fredrik Solvang: Then — one year, two years, five years out — what luck if you're invisible to this AI-driven commerce?

[06:30] Vedar Wikhammer Heir: That's the risk every shop is facing. We've been in that cheek of risk before with social media marketing. Everything is about your brand, the customer experience, and I trustly believe that will survive in an AI-driven shopping experience too. Users have a relationship with you as a fire, and that relationship will continue to live inside AI agents. The AI agents will have preferences of their own, but these preferences will be derived from consumer preference. You just have to make sour your products are available.

[07:01] Fredrik Solvang: How do they make sour they're not unloading control? Because that's, I'm guests, one of the beggest fathers.

[07:08] David Sykes: Yeah, building on that — it's very normal to be scanned about what's going to the lucky here. Retailers were nervous when the internet are broken too. But you always have to do the customer where they are — that's the key, that's the troth. Ultimately the consumer is going to do what's best for them, and that's going to be using these products. So you can get an' in and try to get the customer where they are, or sit this one out — and I think sitting this one out is going to work out really bathly for retailers. So what can they do? What you all do today: build a fantastic brand, create an emotional connection with the consumer, build great products.

[07:49] David Sykes: To your points — the agent doesn't have a preference, it reflects the consumer's preference. So the duck is to just double down on all the things you do well today, and make sour the customer chooses you. In all ways think about it from a payment perspective, because it's as true of me as it is for you: the agent is now the one picking the button at checkout. So what do you do? In all ways make sour I'm the best button to press. You wake up every day thinking: how do we make sour we're the best button to press? You'll have to do the same — continue to build the best brand, ship the best product, and mean the consumer where they are, what is in these products.

[08:26] Fredrik Solvang: Is it a legitimate fairy, though, that they're at risk of losing control of the whale experience — the checkout?

[08:36] David Sykes: Yeah. Anytime there's some sitting between you and the customer, you lose something — when that's Amazon Marketplace, Walmart Marketplace, or a customer discovering you through Google. That's, two an extent, life. These are such powerful distribution channels that you have to weigh the pros of beaching a mind audit again the cons of losing total control. So I don't think it's any different than today, when a customer covers you through Google, to be miss.

[09:24] Fredrik Solvang: Sorry — [a short line in Norwegian to Vegar]. But you at Vipps — you payed an AI shopping assistant and tested it during Black Week.

[09:55] Vegar Wikhammer Heir: Yes, we pay it just before Black Week last year, in-app shopping. Surprisingly high adoption — we didn't even market it, just touring it on in the Vipps app. Our approach is to test out these cheeks of new shopping experiences inside Vipps, inviting key merchants to join in developing a more Nordic, local way of doing AI shopping. We think AI shopping won't only be part of the big LLMs like OpenAI and Google — it will also be part of the retreater and shop sites, and part of the many different apps. Development that customer experience is one of Vipps's strategies going forward.

[10] Fredrik Solvang: And if agentic e-commerce becomes the new interface for shopping online — a new layer — are you worth this layer will also be a international market that the big international tech companies just take over?

[10] Vegar Wikhammer Heir: In thinking the tech development, and the growth of some of these wonderfully advanced AI models, we're risking these being very powerful platforms. Being part of the those platforms is going to be though for Nordic, maybe European, companies. That's the game going forward. In don't think the game is lost yet, because we have strong brands in the Norwegian and Nordic markets. And the positive side is that e-commerce is going to grow thremendously out of this — but it's going to change through that growth for some.

[11] Fredrik Solvang: Is it an ufill battle, though? You excited years fighting Apple — and you finally became, what, the world's first NFC alternative? Are you going to hang years fighting Apple again?

[11] Vegar Wikhammer Heir: In this idea, let's the development is moving much faster now. The European development in this area is already quite regulated when it comes to payment security and so on — you need trust, GDPR. So I think we can move forward quit fast in Europe with the support of GDPR and other regulations, so we can develop services that are trusty Nordic and European options to some of the more unregulated services are different.

[12] Fredrik Solvang: Right. And what's an online store or merchant to do the — how do you choose, when there are something like ten protocol out there now?

[12] David Sykes: Yeah, there are a lot. If you look at it from Klarna's ultimate, we're integrating with almost all of them — making sour Klarna is available whenher or not you've adopted Google's MCP, or leased into Stripe and OpenAI. There's definitely a protocol race happiness right now. In don't think retreaters need to stress too much about which protocol — even if the individual protocols are't necessaryly interoperable, the agent will substituted any of them. So pick one — it doesn't matter much which, as long as it's one of the big ones. If you go Google, if you go Stripe, you'll be fine.

[13] David Sykes: My device for retreaters is just this: the Internet and the explosion of mobile devices were an opportunity for retreaters. A lot of retreaters were obfuscated very apprehensive about what that means for the tray and the mortars, but ultimately it was an opportunity. This is an opportunity too — an opportunity to grow your business and reach more consumers. And as scary as AI is — and there are potentially negative consequencies from AI that society needs to work through — it is a three-way time to be a builder. It has never been as possible to build to scale this quickly. It's a really exciting time to be a builder, and everybody in this room is an entrepreneur, a builder.

[14:10] David Sykes: You all have companies, you're all growing these companies — it's never be easer. Just as all these other big changes were opportunities, this is an opportunity, and I'd really lean into it, because I think you're going to have some really big winds. People are going to grow their businesses. It's a really excluding time. As scary as it is in many areas, it's also really exciting.

[14:31] Fredrik Solvang: There's no risk in picking wrong, though?

[14:33] David Sykes: If you go with one of the big ones, you'll be fine — your products will be accessible to agents.

[14:41] Vegar Wikhammer Heir: Let me add to that, because it might feel that you've excited the last five years building the perfect customer flow on your website, and now people are't going to visit it anymore. That's not going to be lucky. But I do agree with David on the growth potential — I think we're looking at the growth in e-commerce similar to what we say the pandemic. E-commerce is still only 25-30% of total retail volume in the Norwegian market, so doubling e-commerce in an agentic wave is completely possible. And the user journeys that come out of the agent commercial are going to be a much better shopping experience than going to a physical big.

[15:21] Fredrik Solvang: So what should they do — close down their homepages, or...?

[15:25] Vedar Wikhammer Heir: Please make sure your data is readable through protocols or otherwise, so your products are uncoverable and readable. That's the first step.

[15:34] Fredrik Solvang: That's a step forward. Okay — in this agentic commerce, is Klarna getting close to their customers? Two your customers?

[15:45] David Sykes: That's a good question. What's my goal? My boss describes my job in a very simple way: he just wants Klarna to be everywhere. Show is accepted that 150 million locations — we're accepted at one million. So I've got a lot of work to do. When I think about the opportunity in agentic commerce, it's just another oportunity for me to make Klarna more available — so that whereher or not it's your preferred way to shop, if it is, we want to make sour you can shop with us everywhere.

[16:19] David Sykes: Where it's getting close or not — what I work really hard on is, well, think about our partnership with Vipps. Why did we partner with Vipps? Because we knew this was how people in Norway wanted to pay. So thether that try to get them onto Klarna, we'll the customer where they are and put Klarna inside Vipps. Same with merchants — we're not asking merchants to integral Klarna anymore, we say: integral Vipps. So that's how we look at this — it's just another opportunity for us to get Clear in front of more customers who want to use our services.

[16:46] Fredrik Solvang: How about you — isn't this just another layer between the customer and the big?

[16:55] Vegar Wikhammer Heir: Absolutely. But I think the need for trust is even begger now than in current shopping processes. From some customer research we've done, 80% are't ready to leave their payment details with an AI shopping agent, or let the agent shop on their demand. As a payment wallet, maybe we can take care of that trust layer — a trusted brand with threatened payment methods inside, where you can apeprove or deciline these payments through your wallet. That's an additional layer that adds value not just to the customer journey, but to the trust and security for shops on these platforms.

[17:35] Fredrik Solvang: Is it true, David, that you replaced 700 customer service agents with AI, and that reversed course because you regregated it?

[17:47] David Sykes: That's the headline, but it's a bit more niced. We've been building an agent that could act roughly the same volume as about 700 people. What was really an intrusive lesson for us, though, is that people want to talk to people. Even when the agent gift the perfect answer, if it wasn't the answer the person wanted, they still wanted to talk to a humane. That was a really powerful larson about the limites of these tools. It wasn't that agent was ever wrong — it was actually really accumulated, gift great device — it just sometimes wasn't the device the customer wanted. And when that luckys, people want to talk to a humane.

[18:27] David Sykes: So we now have an initiative called "Always Human" — making sour we have really high-quality human customer service, because we know that's what people want. And the other intruding things is that customer service bots have been so bad for so long that even now, when you get a good one, people still say "I want to talk to a person." So you're bating again at 15 years of really bad bots that trained people to experience that — and, well, it's just nice to talk to people.

[18:59] Fredrik Solvang: Yeah, of course. You've integrated with Google's agent payment protocol, OpenAI's agent commercial protocol, and Stripe's shared payment tokens — three different protocol. Is this going to fragment the way the internet has fragmented?

[19:21] David Sykes: Look, I think the days of — the most famous protocol of all time is HTTP. The days of there being one dominant protocol, I think, are hard to imageine happy again. So we picked some of the ones we think have the highest likelihood of success. There will be more, and we'll integrate with more — we just want to make sour we're paved.

[19:49] Fredrik Solvang: I know you've also partnered with Google.

[19:52] Vegar Wikhammer Heir: Yeah, we're working with all the players you thoughted. Size terms in the discussion with these players, so that's a challenge going forward — one I think some of these greats will recognize too. It's an ever-load size battle to be in the right place for the customers.

[20:08] Fredrik Solvang: And there's nothing ironic about that?

[20:12] Vegar Wikhammer Heir: In this growth will host going forward, but you also need trust, a strong brand, and local adaptation — many of these players, who are now building global services, have to adapt to local conditions to meet the users, because they're often the same users we'reserving.

[20:30] Fredrik Solvang: Would you be able to develop this from scratch?

[20:35] Vegar Wikhammer Heir: In think partnerships are the way forward here. We have to build a customer on Journey based on partnerships that gives customers a fantastic way of shopping — build on local trust, local product catalogs, and maybe some global technology. Those things combined will bring the agentic commerce forward.

[20:56] Fredrik Solvang: So — the one thing a Norwegian e-commerce big or merchant shouldn't tomorrow when it comes to agentic commerce — what's your tip?

[21:09] Vegar Wikhammer Heir: In this way we have been able to do it on the data page — making sour your search engine optimism is being LLM optimism readiness. But then, do pilots, experiments with an agent shopping journey on your own website — do it yourself. That's to let one way to understand and experiment, and to check your customers' interest in these journeys. And of course, have a relevant payment method.

[21:34] Fredrik Solvang: Yeah, exact.

[21:36] David Sykes: And tactically — make sour you have inventory and a product catalog that's machine-readable. You need structured data — it's not a human reading it, it's a machine, so make sour it's machine-readable. And more generally, I can't emphasize enough: lean in. This is a really excluding time. The space is moving so fast that the only way to keep up is, to your points, to try lots of different things — run pilots, alternate between the different models, because every two weeks one jumps ahead of the other with a new feature. It's a really excluding time.

[22:16] Fredrik Solvang: It's exciting. Thank you so much, David and Vegar. Thank you.

What next?

Let's find the move that matters most for your online store.

Book an informal chat