Fascinating to talk to Kaspar about the future of agentic AI in our business to business models. It’s getting wild. Visit the accompanying page for all the facts, stats, shareable quotes and a transcript as markdown to feed you AI! https://fieldnotesfromtheagenticeconomy.com/dispatch-02.html Full Transcript Katja Forbes (00:13) Hi, I’m Katia Forbes and welcome to Field Notes from the Agentic Economy, an interview series where I talk to interesting humans about this new intelligence layer that’s landing in our businesses and lives. Today I’m talking to Kasper Korjus from Pactum, an Estonian platform that is introducing AI into procurement, into all kinds of B2B engagements, including having AI as an executive actor. Katja Forbes (00:43) Welcome to the conversation, Kaspar. It is so fantastic to have you here. Kaspar Korjus (00:47) Thank you for having me. Katja Forbes (00:49) let’s start at the very the very beginning for people who perhaps haven’t heard about Pactum, haven’t heard about AI acting as a negotiator inside our businesses. Kaspar Korjus (01:04) Yeah. Katja Forbes (01:04) the the things that are happening in business to business conversation in the agentic economy. I guess just just sensitize us to some of the basics, please. Kaspar Korjus (01:13) so Pactum AI established seven years ago, and it’s it’s like digital workforce of agents sitting in enterprise procurement functions. And they they they do negotiations, then they clean data, they they process requisitions, sourcing events. So like any digital work any worker would do in procurement enterprise, these our Pactum agents are doing the same stuff now. Katja Forbes (01:38) Wow. And I know that you’re working with some pretty big organizations. Are you able to share how some of your clients are using these is it fair to call them AI workers or is that just a bridge too far? Kaspar Korjus (01:52) The industry is calling them differently, like some of them labeling with names of humans as well. Like I I don’t like that path. they are still not humans, right? But like Katja Forbes (02:01) No. Kaspar Korjus (02:01) digital digital colleagues, maybe fair to say a digital workforce, right? and yeah, different use cases. How we started was our first client was Walmart actually, the world’s largest Katja Forbes (02:11) Yeah. Kaspar Korjus (02:12) and we started by them just telling that, you know, they have five hundred billion in spend and hundred billion of that they don’t have people. Kaspar Korjus (02:19) To Kaspar Korjus (02:19) manage and so if our agents can kind of reach out to these suppliers and and you know help them and find better deals and and see that all the prices are fair, then you know if you can save one percent out of 100 billion, you save billion annually, right? So it was pretty obvious business case for AI agents to start with this long tail of supply chain. but now what the agents are doing is like end-to-end procurement workflow. let’s say you want Kaspar Korjus (02:46) buy a thousand iPhones. You do a requisition, then you you need to see whether it’s in catalogue. So agent helps to do that. Then it checks that policies are all correct, that whether it’s correct commodity that has chosen and all of that. Usually there are 10-15 people enterprises who kind of check these policies. Then it reaches out suppliers, then it asks whether they are interested to supply them. Then they ask quotes from the suppliers. So then they negotiate with the suppliers where we got started, right? And then we got better deals. Kaspar Korjus (03:16) And then we update prices in the SAP and other systems. And so all of this end-to-end can be fully autonomous with smaller suppliers and with larger suppliers, then humans are in the loop somewhere there. But it’s pretty fascinating to think now, right? That basically majority of the work what procurement functions are doing can be kind of automated. Katja Forbes (03:40) Mm. Kaspar Korjus (03:41) Like it or not, that’s another question, right? But but this is the world how it goes. Kaspar Korjus (03:45) Mm-hmm. Katja Forbes (03:46) Well, yeah, Katja Forbes (03:46) let’s talk about those sort of the like it or not. so the suppliers, d help us understand what would it be like for one of those suppliers to have Walmart’s AI negotiator, AI procurement through the Pactum platform? What would it be like for them? Do they get an email? Do they get a phone call from a robot voice? what would the experience be for those suppliers to work with this machine customer? Kaspar Korjus (04:14) Usually this via emails, and text messages as reminders, chat interfaces, so the channels can be different. But yes, as a supplier you’re reached, let’s say, by an agent and asks, you know, whether you want to make bid for this this item or this event and then basically goes through and and suppliers actually like it. First of all, usually they get more business that otherwise they might have not. Second it’s very fast and they can, you know, they can take their own time to do it. so Katja Forbes (04:43) So how fast Katja Forbes (04:43) is fast? Let’s go. Like how would how long would it usually take and then how long does it take now that we’re in this fast mode? Kaspar Korjus (04:49) When it’s about negotiations, it’s usually ten fifteen minutes of exchanging terms, understanding, making offers, counter offers, while with humans it’s like several weeks of back and forth. Katja Forbes (04:58) right, okay, so Katja Forbes (05:00) we’re going down from weeks to minutes. Kaspar Korjus (05:02) Yeah. Kaspar Korjus (05:02) And this is only negotiation part when it’s actually creating sourcing event or checking requisition. Only checking requisition is usually five day SLA, meaning that it takes humans five days. For agent it takes thirty seconds. So basically if you want to buy ten thousand iPhones, then instead of waiting for five days for some procurement professional to say whether you can do it, and then comes back that no you can’t do it like that. Now it’s like after thirty seconds you can actually purchase that. So this is the velocity and speed what basically agents allow to work twenty-four seven. Katja Forbes (05:31) That’s wild. And I just read the release that you put out, your requisition alignment agent passed one million requisition checks since March, four thousand times faster than manual reviews. Kaspar Korjus (05:39) Yes. Kaspar Korjus (05:43) Yes, yes. And and this is maybe the scary part, but basically at the same time it’s the part but not what humans don’t want to do. Like no one wants to check tens of thousands of requisition line items. Like I don’t know, our human minds are meant for something greater for that. So I don’t I don’t see that people are like disturbed that the agents are doing that type of work now. Kaspar Korjus (06:03) Of course we don’t Katja Forbes (06:04) Hmm. Kaspar Korjus (06:04) know what happens in two, three years’ time because the speed of the models and it’s so so exponential so our minds are not capable of even thinking on that exponentiality, right? That once these at the moment the agents are acting on kind of operational layer, but now basically our agents are already listening in strategic discussion. Kaspar Korjus (06:24) Discussions and management meetings. There they are not yet participating, meaning that they don’t kind of come and interrupt you and say that, hey, you know, Katja, you’re wrong here. Actually, you should expand to this market because of these trends and stuff, and look at this competitor is already doing that. They are not there yet, but it’s there definitely is like a matter of years before they can enter the strategic discussions. But there, I still think that so I’m quite optimistic in that sense, that I still think that they will be colleagues for us. Kaspar Korjus (06:54) They are not going to replace the strategic thinking because the way how large language models today are are built. I I don’t believe that they can take over all the jobs of humanity. Katja Forbes (07:06) I don’t think so either. and I there’s a couple of threads I wanna pull on there. One of the threads is what you’re seeing inside your clients and how they are changing with these new digital workers. so maybe we’ll pull that thread first and then I’m gonna come pull that other thread on the the the strategic advisor inside the executive layer. Kaspar Korjus (07:33) It’s clearly the most difficult part. It’s not about technology. Technology capabilities are are quite enormous there, but it’s the change management part. And how how do I feel different this time compared to like software as a service layer or some other layers? It was that at least in procurement the function was always least developed because you know it wasn’t the most important function compared to sales, let’s be honest, right? Katja Forbes (07:58) Mm-hmm. Kaspar Korjus (07:58) some other functions. So so but now with the agents Kaspar Korjus (08:03) there is such a great energy that people know that this is not going to last, what we have built for last decades. And the speed and kind of space how they want to change now by chief procurement officer level is insane and very supportive in that sense. And everyone knows that this is going to happen. So now with agents, actually, it is easier to kind of actually transform these enterprises. And this is this also means kind of looking through what roles of humans we still need, where our agents where Kaspar Korjus (08:33) work together and kind of making this transition together. And and it’s not clear. Like it’s it’s still discovery of like in in indirect in some phases it’s clear that we can we don’t need humans at all because humans are just time and doesn’t add value. But in strategic discussions and and category strategy and stuff there is
