Summary

In this episode, Alp Uguray interviews Kaspar from Pactum about revolutionizing negotiations with AI. They discuss Kaspar's background and the founding of the e-residency program in Estonia. They also explore the concept of negotiations and how AI can assist in creating value for both parties. The conversation delves into the future of negotiations, the transparency of AI decisions, and the parallels between entrepreneurship and participating in an Ironman race. Kaspar shares insights on decision-making, prioritization, and the importance of mental resilience. In this conversation, Alp Uguray and Kaspar discuss the potential applications of negotiation AI and the challenges of implementing it in various industries. They also touch on the topic of AI regulations and the balance between innovation and bureaucracy. Kaspar introduces Pactum's four-level negotiation intelligence framework, which categorizes AI negotiation use cases based on complexity and impact. They also discuss the importance of building a foundation layer for AI in government services. The conversation concludes with a discussion on the benefits of Estonia's e-residency program for startups.

Key Takeaways

• Negotiations are happening everywhere in life, from personal relationships to business deals.

• AI can assist in negotiations by understanding potential terms, their value, and making trade-offs.

• The future of negotiations involves both parties using AI agents to engage in continuous negotiations.

• Transparency in AI decisions is important, and interfaces need to be developed to provide clarity and control.

• Entrepreneurship and participating in an Ironman race require mental resilience, quick decision-making, and the ability to prioritize tasks.

• In startups, it's important to postpone decisions that can be made later to gather more information and make smarter choices. Government contracts offer long-term and substantial opportunities for AI implementation.

• AI negotiations can lead to competitive advantages and market dominance.

• Regulations play a crucial role in ensuring transparency and accountability in AI negotiations.

• The four-level negotiation intelligence framework categorizes AI negotiation use cases based on complexity and impact.

• Building a foundation layer for AI in government services can lead to more efficient and innovative public services.

• Estonia's e-residency program provides time and cost-saving benefits for startups.

Episode Chapters

• 00:00 Introduction and Background

• 03:04 The Concept of Negotiations and AI

• 04:52 The Future of Negotiations: AI Agents on Both Sides

• 08:01 Transparency and Control in AI Decisions

• 11:08 Parallels Between Entrepreneurship and Participating in an Ironman Race

• 15:55 Prioritization and Quick Decision-Making in Startups

• 28:05 The Potential of Negotiation AI in Government Contracts

• 30:28 Navigating the Challenges of AI Regulations

• 42:21 The Four Levels of AI Negotiation Intelligence

• 48:15 Building a Foundation Layer for AI in Government Services

• 52:33 The Benefits of Estonia's E-Residency Program for Startups

Transcript

Alp Uguray (00:10)

Hi everyone, welcome to Masters of Automation. Today I have the pleasure of hosting Casper from PaKtom. Casper, thanks for joining.

Kaspar (00:20)

Hi, nice to be here.

Alp Uguray (00:22)

So first of all, thanks to Jake for helping me organize this episode. I wanted to go right into it. mean, you have a stellar career for starting the e -residency program in Estonia and now Pactum and you just closed your founding round and really revolutionizing the negotiations with AI. So I have some questions there. But before we dive into all that stuff,

What was Casper like before everything began?

Kaspar (00:57)

That's it.

I guess I'm the same person still, just living in countryside, in nature, with family. Fishing, just came from Lapland, fishing trip with my brother. Four days of heavy fishing. I don't know which was more difficult, being four days with my brother or four days of fishing, but exhausted today.

and always having in mind kind of realized that the university or something that, we can shape the world around us as we want to, that the world around us is being built by someone not smarter than we are, it's quite hectic and random, so we can just imagine where and how it could be and then just try to change it. And quite often, if you're persistent, it may work.

Alp Uguray (01:53)

Yeah, and I think the, especially before, for example, you got started with the residency program, what were you up to? Like you were in school and then you were like thinking about a way that we need to change things here. What was your trigger?

Kaspar (02:12)

The trigger there was that government asked the question, what if we have 10 million Estonians by 2025, so by end of next year. And asked the question by, and then I took this as a scholarship from government and started to kind of build it out. And honestly, like through time you get

you understand to some vision of, okay, we could have governments digitally without borders, okay, this could be use case and okay, this could be eventually we lead to the future of nation states, of borderless nations, nation as a service. And this builds up like, of course, there are some entrepreneurs that really work on moments in their lives. But for me, it's mostly through work, you get this vision and you adapt and you improve the vision.

Alp Uguray (03:04)

So as you work on it, you start seeing the areas where...

Kaspar (03:08)

Yeah, and later you can be smarter about how you had this idea from day one. But usually it grows and it builds together with you.

Alp Uguray (03:19)

And now it's a massive initiative. I've been exploring about the residency program as well. Actually, a year back, Stonehowl, the program is set up and it enabled a lot of entrepreneurship in Europe. It enabled lot of, it removed the barriers of entry, I think, for a lot of people. What were some of your, while you were working on it, like you said, right, like keep seeing the areas of...

What led you to see the delight to pactum while you're looking at those opportunities?

Kaspar (03:58)

It was together with two other founders, my older brother, Christian and friend Martin, who brought the topic of negotiations to our table. And for me, negotiations was like, I knew it happens, but I didn't know much about that. we went into science first, we went into science and realized how much value is left on that.

because we are bad in negotiations. And then we realized negotiations are happening everywhere. And it was like this matrix, neo moment when things turned to green. When you see that, my kids are negotiating with me. I need to negotiate with wife who takes the, I don't know, child on Thursday, right? I need to negotiate with the hiring, with clients, with sales. Like a negotiation should happen everywhere. And there's so much value if you understand that you can create value to other side by just try to understand what they like.

try to extend terms and then get something in return, get better contract and better terms in return. And when we realize this, that they are happening everywhere and people are bad in that, then the question is only whether AI can solve it. And of course it was before the chat GPT era, like few years before we started that, then chatbot type of interface is very too sexy, right? So it was like more support system. But our bet was on chatbots.

and AI to be an independent agent to do negotiations for humans. And five years later, it seems that it wasn't a bad bet.

Alp Uguray (05:31)

Yeah, I like how you said like negotiation is everywhere in life, like even at home to you, like at work to different verticals as well. How do you see the like the first users of the platform, especially when it comes to like, okay, I could negotiate and then and then maybe decide at the end, reach out to decision. How is that decision tree look like, like from from the perspective of AI?

And then one day I see the contract and I'm like, wait, AI negotiated this well and signed on it too. How does that part?

Kaspar (06:08)

Yeah. Yeah. So for the AI to have authority, it needs to know some things. First, it needs to know what are the potential terms that can be negotiated. So they are fixed like price, like payment terms, like stuff like that. Second, it needs to know the value of those terms. So it needs to know that if I exchange early payment, if I pay you 30 days earlier,

Can you give me 1 % discount? So this is equal value to me. And then if it knows the terms and the value, can start making trade -offs. And when you have 10 plus terms, it's quite difficult for human to even calculate those trade -offs. If I would ask you, are you ready to lower the item cost for these five items 2 .5 % while paying me 30 days earlier and changing the warranty of these terms 30 days later?

while actually also giving exclusivity on the other term. It's some math problem. For AI it's easy. It's like Excel. You have rows of data and values and it can do the math, right? These days already pretty well. And then the third element is, okay, AI knows terms, AI knows value. Now we need to exchange information. And email is not the best way to do it because it's too open -ended. Discussion can go to random things. And so...

And it takes time. So chat was our bet there and structured chat so that some parts of the chat, it's all structured. It goes through introducing themselves. The AI says hello to you. It asks your preferences. Then it goes to terms, et cetera. And some parts are then open -ended and you can do offers. And so three main elements there and the AI, the flow as you asked is that usually we try to increase the pie. So first after introductions of what's going on, then we try to ask

supplier preferences, where the supplier wants to grow, what problems they have, and then offer these services to them. And in return, then people want to offer something in back. And then we ask these things in return what we want.

Alp Uguray (08:15)

What was the toughest negotiation that you've seen AI do?

Kaspar (08:21)

Yeah, it's the negotiation. Our pack to Mea is executed in mass negotiations where there are very repetitive negotiations that are more easy to automate when we are here at Masters of Automation podcast. It's like things that happen regularly is potentially automated. If one -off, big thing, then you better do it once by humans, right?

And so our product applies to these cases, but of course we have had cases where there have been contracts worth of tens of millions that the AI goes and negotiates and gets some better deals.

Alp Uguray (09:06)

So then it becomes very valuable if the contract is very large and then in a way that it removes the step of like having an accountant and a lawyer reading and then trying to figure out like what are the certain terminology is and identifying those areas and then keep pushing for the benefit of the company. Is that like how you're seeing the platform?

Kaspar (09:30)

Where AI can assist humans? when it comes to AI, it doesn't have that human intuitive mind still of knowing about terms that haven't been discussed before or agreed before. So there AI can suggest negotiation strategies for humans and stuff. Where back -to -me AI and where automation makes sense are in occasions that happen like

Alp Uguray (09:34)

Mm -hmm.

Kaspar (09:59)

hundreds and thousands of times annually, like purchase requisitions, like contract renewals with 20 ,000 suppliers, like big sourcing events that happen every day to buy the same stuff for every month or every week, and these type of negotiations, repetitive ones, logistics truck lanes, know, Maersk needs to have tens of thousands of them every month because prices are changing. But when it comes to &A deal,

between Meta and another Snapchat, then you wouldn't send back to Meja a negotiation to do that price negotiation. Then it's complex of teams and humans needed today.

Alp Uguray (10:41)

Now that makes a lot of sense. In the future where both sides of the parties in negotiation table use Pactum or an AI agent to engage in a continuous negotiation, is that kind of the future that you see as the product evolve and as actually industry evolve as well? Because both sides of the party adopt an AI agent.

Kaspar (11:08)

Yes, and this is very interesting and there are different interfaces for that. But definitely what we see is that first people and businesses start to realize that big part of their sales is done by AI bots, that AI bots are buying stuff from them. Like it can be Alexa who buys dog food from a dog food shop.

because it became a trendy thing that Alexa discovered that, hey, your dog doesn't have any food left because dog has eaten already 20 times. And last time you purchased the 25 times. So it's five times left. So we need to order. Hey, do know that? And then when this becomes practice, then Alexa becomes the buyer for that dog food company. And then basically the dog food company needs to adapt their sales channels so that Alexa would eat

could easily order this food, right? To go to their website and log in and do users is very human thing to do. So they need most likely some API connection between Alexa to sell this dog food. So, and then basically negotiation starts happening there as well. Hey, Alexa can ask, can I order 10X? What's the price then?

And the API should enable to do that. And then basically it's API to API connection negotiation. And you as a dog owner don't care at all. You have given budget and authority to Alexa that please feed my dog. You have 100 euros monthly. And then Alexa can go and source best dog food and get best discounts for your dog. And you don't care. You don't need to put time there. And so this definitely can and will happen.

sooner or later to have additional AI as a sales channel for both sides and then websites and different parties need to adapt to this and we have already seen that also at Bactom for the first desires.

Alp Uguray (13:15)

How do you see the AI transparency layer evolve over time? For example, speaking of Alexa, Salt Park in one of their episodes did an experiment a few years back. In the episode, Eric Cartman was saying, Alexa, go ahead and order me 10 ,000 toilet papers. So the case was everyone's Alexa at home got activated.

and then placed an order for 10 ,000 toilet papers, including mine because I was watching the episode as well. And then it went ahead and it was placing the order and I canceled it. It's a true story. How do you see the, that AI transparency layer evolve so that we see full clarity of like the decisions that the AI agent makes and also the

Kaspar (13:56)

Thank

Alp Uguray (14:10)

the trackability to build the guardrails.

Kaspar (14:13)

Yeah, yeah. And with every new technology, it first strikes to places where we couldn't imagine it, right? We couldn't imagine that in South Park, they order that and these things happen and then we react to that and then we build on top of that, right? So, I think...

giving transparency to AI decisions is a definite must to a certain degree at least because as we now will have tens and maybe hundreds of agents independently authorized to act on your behalf not just to chat but act and have decision power most likely also budget to some extent

Then yes, human needs to understand where money is spent, what is done. But technically, I don't see a problem for that. We just need some interfaces. And mostly, it will be controlling the UI to limits, to trade -offs, drain your agents to act better for you, give feedback to the agent, and kind of...

And this is what our clients are doing also. Our clients that you can read from, don't know, you can read from web is like Walmarts of the world. Their users, their buyers are giving limits and input to our AI agents to do the work that they would do otherwise. So basically, so every person becomes like a manager of hundreds of agents that coexist with you.

and now at their workspace. so the work changes, like that you need to now, instead of doing yourself, you train and you put limits and thresholds to AI agents to act on your behalf. And that's fascinating because it really empowers us. can do much more with much less time.

Alp Uguray (16:16)

And it enables knowledge as well, because not everyone is trained to be a great and or fierce negotiator. So like in different contexts, I'm able to choose my risk appetite, my budget and like the risk that I am willing to take in the negotiation table, like even a poker game. I think that that really enables and democratizes access to the negotiating power on the table.

Kaspar (16:34)

Yeah.

Exactly. And sometimes just like there's so much in life that we need to do like and you don't have time to deal with everything. And so which means that if you're a good person, if you're good strategically and you're successful, you're prioritizing your tasks all the time. You do the 80 -20, you do these tasks that matter the most. And then your wife is upset because the door is not still fixed.

for a month. But the is not my top priority right? But someone needs to reach out to, I don't know, door fixers. Make a shit the price. Ask them to come here and fix the fucking door. Right? And we have so much with our animals, with our schools, with work, with children, with our personal life, with sports, that all the time needs something that also not just chats, but sometimes has authority and budget and to do stuff.

Alp Uguray (17:13)

Yeah.

Kaspar (17:39)

So it's not the only skills of negotiations just helping us as human beings to cope with our lives that are so complex these days.

Alp Uguray (17:47)

It's like it's also time element to it. It's like who's going to drop the kids to school today and then two AI agents start to negotiate.

Kaspar (17:54)

And you can have some quality time and wine with your wife and husband, right? And they are dealing with who takes the children part.

Alp Uguray (18:04)

That would be awesome.

Kaspar (18:08)

Maybe it would save some marriages.

Alp Uguray (18:10)

Maybe, exactly. I'm curious, Jake mentioned that you are an Iron Man or you're preparing for an Iron Man.

Kaspar (18:22)

No, Jake has done some good sales man, sales job, But... In all honesty, it was Ironman73, which means that this is the half of the Ironman, so... So I'm half Iron.

Alp Uguray (18:25)

It's all you great

That's still, some part iron is good. So I'm curious while you're preparing for that, mean, Ironman races are intense. It takes grit, determination, Like hating to do something, but also loving it. What are some things that feel similar in entrepreneurship and being an Ironman?

Kaspar (19:07)

It's a good question. It's a good LinkedIn post. These days everyone posts things on LinkedIn privately. Trying to connect the dots somehow. Well, one thing is, mostly it's mentally to understand that this is your mental, it's a mental problem, not physical problem.

Alp Uguray (19:09)

It's a collective thought.

Yeah.

Kaspar (19:31)

that you can survive this and you can reach next phases if you're mentally wanting to do that. And in startups, always basically, you're in trouble. You always have this paranoia of things are bad. And then mentally just driving through there that despite no matter what basically, you will get to the next phase and next phase and next phase and making the phases short not...

becoming unicorn on next day, but just understanding can I achieve this in next week or month? And then if it milestones are shorter, like in Ironman, you divide this big challenge into shorter wins, then it's easier to achieve. But yeah, there are many, parallels, I think. Follow me on LinkedIn and I'll make a post.

Alp Uguray (20:21)

Yeah, that's a great link. You should make a LinkedIn post right after this. What are some things though, like from your experience that led you to be successful, like just building daily habits, like we're like planning from product perspective, like, because there are many customers and then they all want different things or a lot of things. How to manage that.

Kaspar (20:25)

you.

I don't know whether I succeeded that. It's difficult to say no. Mostly it means saying no, right? But you can't say no too early because you're too stupid too early. So you some training goal with some data points. And then when very limited amount of data to start saying no and pray again that you can change it back, revert it if it was a bad decision. But I guess quick decision making.

saying no learning, adapting is the same kind of small wins going back to small wins like making bigger milestones through smaller wins and then trying to appreciate these smaller wins and business.

Alp Uguray (21:37)

And then have the ability, think what you said is I have the ability to make a decision, but also have the room to make more decisions and like with a flexibility or as a startup.

Kaspar (21:48)

Yeah, guess one of the like it's it's one of my favorite parts of the lean startup theory for startups. I was quite fond of that like 15, 20 years ago was that two decisions that you can do tomorrow if resumptuously, if you can do tomorrow, if you don't need to them today, because in Estonia, there is this sentence of opposite. Don't put the

tasks of today to tomorrow or something, do them today, something very traditional kind of Estonian thinking that... But lean start up is other way around. Like if you don't need to do today, do it tomorrow, because tomorrow you're smarter. And this kind of always to remind that not to jump on conclusions kind of too early and then detecting when it's not anymore.

time and now you need to do decision making. think this is always the balance of everyday balancing of decision making.

Alp Uguray (22:47)

I think the Estonian phrase that means like do not procrastinate, whereas the lean startup is like procrastinate, but keep what is important or because tomorrow you'll be better suited to make that decision because you'll have more experience.

Kaspar (23:06)

Because many decisions don't need to be made today. Many decisions can be postponed at least a day, if not a month, sometimes weeks or months. Yeah.

Alp Uguray (23:15)

Or what are some decisions that you would say you're postponing today?

Kaspar (23:22)

No, many many like

When in work, let's postponing the decision. When are we going to double down in direct logistics negotiations use case? Like when we are building out big team for that, we have success stories with clients. Do I need to do that decision today? No, I can do it tomorrow at least because our sales leads in that space are not that far yet. And after a week.

You're so much smarter how the product can be part of our standard product portfolio. What are the leads? What market tells us? What features we need to build? What we already now have after that week? So all of this in a week time, I'm much smarter. But I'm asked today before this, I had a meeting and I was asked the date. And I said that I don't tell the date today. I don't need to date the date. I can say the date some weeks later. So let's come back and tell him the date.

some weeks later when we actually need to tell.

Alp Uguray (24:22)

taking the right moment. And I thought at one point you were looking at the door and then you were going to say the door, I need to fix the doors.

Kaspar (24:33)

Of course. Then most likely if I'm fixing it's already too late. My marriage can be broken before so these decisions can't be postponed anymore.

Alp Uguray (24:39)

you

But it's an interesting point because building, like having the use case and then having the customer success stories and then building the product for it, especially negotiation and then different verticals as quite a large impact. Like I think there's negotiations in HR as well, right? Like there's negotiations in manufacturing to trading to even healthcare.

like insurance providers negotiating. What is the number one thing right now that you are deciding today that, okay, this is the thing we're doubling down on?

Kaspar (25:26)

Yeah. So we have, like you said, we started also from HR negotiations to sales negotiations, hiring negotiations were very cool.

So we have tried different things, even tried to, I mean, we had a panel with the president of Estonia and Jan Tajlin about government to government negotiations about, I don't know, CO2 emissions between trades or something like that. So Brexit negotiation. But we have doubled down now into procurement negotiations and ICP ideal customer profile is a manufacturing client.

who has lots of C parts, lots of parts they use in production. So this is for the direct use case, the products they purchase for themselves to build something for a production line. And then in indirect products they use for themselves, like I don't know, IT services or laundry services or whatever HR service. Then this is indirect in procurement and for indirect we have, yeah.

use cases, the sourcing events, the purchases, the single of purchases and renewal of contracts. Even this and then of course after that comes logistics direct use case, retail direct. So it's a pipeline roadmap of different uses that come after that. But it's interesting to find kind of when do you stop narrowing down the business and when you start extending it again. Because so far throughout five years we are fully narrowed down.

always become more focused, more focused, more focused. And eventually if you start kind of really scaling in that one narrower use case, then it's time to extend right again. But it's very kind of difficult because narrowing means basically saying no in hope that your product becomes better tomorrow than right. In hope that there are more customers in narrower focus and more revenue earned there than saying no at the moment.

to other opportunities. And this is very difficult because there are obviously with growth companies, there lots of revenue expectations from investors and everyone. And then saying no to revenue, it's difficult.

Alp Uguray (27:36)

And I think that's like some of the parts of slippery slope parts of the lean startup, because in lean startup, it's like keep narrowing down on one niche and then like, dominate that market or that niche and then grow and expand. then there's like with your type of product with like negotiations, like you said, it's everywhere, right? And then there's the opportunity cost of which one to focus on.

Like you said, government, government contracts are pretty long, like 20 -year contracts and large substantial contracts if implemented. So there are a lot of benefits there as well. So if you were to think of the opportunity cost from the perspective of lean startup, and you already mentioned, the navigating through it is tough, but...

Where would Casper want to apply negotiation AI if opportunity cost was not a problem?

Kaspar (28:44)

If opportunity goes to be a problem, I would like every person of us would have Back to MEI in the pockets to kind of help us with our daily lives, from shopping to selling stuff, being like super agent working for us. In a bit shorter time frame, of course, the entire purchasing slash procurement divisions we see Back to MEI.

full domination across the globe. And the question then is that are we then too late to take over sales negotiations as well? Then Bactam would control all entire inflows and outflows of a company. But next four to seven years at least procurement negotiations across industries in indirect and direct use case.

Alp Uguray (29:35)

I'm curious from the perspective of AI agent to AI agent negotiation, like there will be other incumbents that build different AI agents, but in a competitive landscape over time, as let's say, Pactum AI competing fiercely with other AI companies, AI agent on the negotiation table, and then beats the other agent by signing a better deal and then better...

better requirements sheet. How does that communicate it to everyone on the table that is negotiating? So do people know that it's an AI that negotiated the deal and won against the other's AI? How does it work?

Kaspar (30:28)

I guess regulations come in play here as well. Like the EU AI Act very clearly states that AI needs to be clear that it's an AI, not a human. But as we know also, EU is quite often overly regulating, right? So the question is what the industry standard will be and most likely...

Most likely we won't see these regulations in the US at least. And which means that, yeah, if you receive a call or email when someone wants to buy or sell something, then you can be never sure that it's not an AI agent. These days, they can't fool anymore in that sense. And of course, when it comes to cybercrime, it's like terrible.

All of a your mother calls and says that she's in a car accident with a taxi driver and you need to send a thousand euros to the taxi driver otherwise she can't get home and cries and then sends the details. Like how to make sure that this was your mother, I don't know. Most likely industry and founders are creative enough and the good wins the evil and we figured out many unanswered questions.

Alp Uguray (31:45)

And then like a lot of, people who are elderly, right? Like face it, face it the most, like who received the call from an AI agent that sells them and then they, then they face that. So like in the balance of, like you said, in the U S it's pretty lenient right now. And then these types of things are happening and in Europe, the UAI act, I think reinforces or like maybe secures, maybe mitigate some risks. But in the meantime,

also restrains innovation and then the way that it is done. In which ways do you see the benefit of regulation today and cons of it in EU?

Kaspar (32:28)

It's a good question, especially because I've worked for a government as well, right? And worked in the times of crypto, where eventually we had kind of the opposite, that I was really supportive of regulations, while the entire crypto community was kind of built anti -regulations. I was built to get out of the government and stuff. And then I was mixing regulations with crypto and trying to launch S -coin, our national cryptocurrency.

and and stuff and why and eventually what it led to that Estonia was by far number one country for cryptocurrency companies and general crypto companies established themselves while quite high fees and the reason was that it helped to understand the clarity for future if I do something X today can I go to prison because it's not regulated yet and later they say that you did wrongly because you know but you know there weren't any regulations

But who cares, right? You did wrong. And so regulations help with clarity. I like them for that. The question just is where you draw the line and where can you kind of empower entrepreneurs and stuff, punish them and increase the bureaucracy. Like you don't think true sometimes when you build regulations and you wonder how much bureaucracy it means for the entire economy. If they need to hire a lawyer,

to read what is EU law and after reading understand where they fit and what it means to their clients and suppliers and how to comply. It's mostly just hiring legal to comply because everyone figures it out how to comply. It's just paying legal fees to do that. And then the EU companies now are paying these legal fees. So most lawyers usually win.

Alp Uguray (34:18)

And then the winner in the deal becomes the lawyers because also startups don't have a lot of money in the beginning as well to be able to afford.

Kaspar (34:26)

Yeah, startups, the EU regulation and going against that is not startup top 100 risk of becoming bust. The product market fits and everything else is a risk. So what startup is doing is that they're doing nothing, right? Because once they're becoming very big and influential and stable, then of course, to manage that risk, you start hiring lawyers and mitigating these risks. but you know,

Maybe that's also reason we don't have too many very large technology companies in Europe at all.

Alp Uguray (35:02)

Yeah, that could be the prohibitive factor of starting something new because the beginning costs are also too high, but also adopting new technology, building the transparency layer of what happened, why happened is huge. I think what were some of the things that you learned from crypto times, especially also including the regulation that now you see

is repeating or not repeating in the AI world.

Kaspar (35:36)

Very good question. I haven't even thought about that. Besides that, is it the same hype as in crypto, where eventually lots of money was poured into crypto entrepreneurs, but not much value was kind of, not much value didn't come out from that, let's be honest, at the moment, at least. And is it the same happening in AI or not in, I don't know, fourth hype cycle, historically, right?

So far, all the hype cycles, I think have paid off through time. It just takes 20 to 30 years. And crypto in that sense is still too new thing to exactly know whether it paid off. Definitely some taxi drivers money went to engineers, which in some sense, I don't know, maybe it's not that bad. Maybe it is. So we don't know that yet about the AI.

But maybe in terms of regulations, there is something to learn that we haven't learned in Europe, that understand where to regulate, what to regulate and empower entrepreneurs always. Entrepreneurs is your economy, kind of. There are criminals, but when you get too stuck with few criminals, you ruin the entire economy.

and you always work with criminals but work with entrepreneurs kind of.

Alp Uguray (36:57)

And it's in a way the problem of having a one that or maybe having a one aggressive student in the classroom, then everyone gets punished for that student's behavior because teacher is like, if he or she does that, you are all punished. So in a way that there's that aspect coming in. And I think in crypto, there's an interesting part as well, like other than

the wallets that are inaccessible, like how much money in it, and then their transactions, cold wallets. The transactions that happen are public, we were able to still see what's going on in the chain. But with AI, think there's still room of the black boxes, like how does it decide, what it decides, and then how did it get predicted? And then there are a lot of different AIs.

AI models at least. So I think that part is quite interesting, especially in relation to open source versus closed search as well. are some of your thoughts around like an open AI type of a model versus more on how Meta is doing? I Meta is not open source, but they gave their model out.

Kaspar (38:19)

I mean, it's more open source than open AI most likely.

Alp Uguray (38:21)

Yes

Kaspar (38:24)

But

There are many influencers for me personally when I'm thinking about regulations and very bad scenarios of our lives through AI. One of them definitely from one end is our first investor, Jan Tielin, founding engineer of Skype, who has been doing research in the fields of

Singularity kind of for 20 years. So nothing new for him. One thing is that in the the community of scientists in this domain and the expected time and things crash for humans. Won't is not postponed, but is coming sooner and sooner through time. It's usually is not happening with predictions. So and one of the reasons there is this race.

right and black box effect of launching and learning and adapting and launching again learning again without understanding and the product is built without understanding exactly how it operates and and what it can cause and launching now that as an open platform for everyone to try things out is things we can reflect in tens or maybe

I mean, it's after some decades or maybe some days, some weeks time, kind of, what's the effect of this? So people who are afraid of these things can be laughed at, you know, most likely throughout time if things don't go bad, but if things turn to bad, then no one laughs at them, then most likely they get punished by they were more serious. So I do support.

all in all, kind of more open discussion about release of code when the harm can be huge and how to manage different risks there. And I do encourage kind of governments and I've done myself in the EU level more kind of to discuss the balance of budgets when it comes to AI safety and AI innovation outside of safety.

kind of at the moment it doesn't seem to be maybe where we would like it to be.

Alp Uguray (40:51)

Absolutely. I think the especially similar thing happened to social media as well before crypto. And then it had its own goods and the bads. I wanted to ask you because PactoMEI developed the four level negotiation intelligence framework. Can you elaborate like how those four levels are? And then the other part is

How did you come up with the four layers?

Kaspar (41:24)

Yes, so what four levels means is that as we have done so many different use cases of AI negotiations, then we have divided these use cases into four levels, level one being standard, then advanced and expert, and then superhuman level. And the four levels, each level has higher cumulative impact of the use case and negotiation and come with a higher complexity.

And level one starts with the most standard purchase, orders, negotiations that could be called a hacklepot even that I want to buy this with that price. Now, okay, let's do it in that price kind of not much diligence built in it, achieve some results. Level two is going more advanced, more terms, more data points, expert level already, which the best humans are capable of doing. Like you have

In retail, for example, you need to understand the margins, commodity costs, the competitor pricing per item level constantly to have good argumentation data levels and negotiation strategies. Very complex kind of and future prediction forecasting as well that AI is ready to and able to do in Bactam already. What we haven't seen too much yet also by Bactam side is what we see in level four, which we call a superhuman, which basically already

good output any human being in achieving results. For example, we have done a pilot in, let's say, logistics use case where there are carriers and then there are shippers and today they are communicating through WhatsApps and different platforms to find routes. So now if you released LLMAI there, who understands the context of tens of thousands of shippers and carriers in single second that never sleep,

Then they reach out to the other side to find whether there is a deal. And with milliseconds, the other side can offer their counter offer to their AI agents and basically transform the entire industry so that deal making is done between machines and those companies who have deployed machines in the market because humans are too slow, like in trading in some cases already they are. And in this case is kind of AI agents, what we call superhuman level.

Indeed, could change today how we operate.

Alp Uguray (43:54)

And it brings so much competitive advantage like to your point, right? Like they can easily turn into dual police or monopolies by getting the market share because they're able to negotiate the best price for the best deal and keep being in profit. Yeah, that's very interesting. So the last part of the podcast, will...

I will ask a few questions if you're okay with it and then small questions and then you can just hit me back on whatever comes to your mind. So what was your favorite book in your childhood and now what's your favorite book now and how they're different?

Kaspar (44:38)

My childhood favorite book and now favorite book. I think my childhood favorite book. need to think. me seconds.

My childhood favorite book was about two bacteria in our mouth who built houses on our teeth. And then they had a great time, lots of sweet candies and stuff until they went to doctor and repaired all the teeth. So it was my favorite book I read end to end and I loved it.

And it was great because it told me that I need to brush my teeth.

Alp Uguray (45:23)

Is it like a book for child in Estonia?

Kaspar (45:33)

Yes, it's Estonian children book, but I'm not sure whether it's translated or it's original Estonian. I need to google that a bit.

Alp Uguray (45:43)

And what about now? What book has been keeping you busy these days?

Kaspar (45:50)

Yes, so my favorite book from last week's has been published by Karpner. It's called When Machines Become Customers, exactly on the same topic what we discussed here. So whoever is interested to read it and you can reach me out and I might find some copies. If not, then go to Google and find it. It's from

don't shape and craft. But then machines become customers and it tells the story of this future where we need to adapt our businesses and lives for the new kind of channel and start to learn to coexist with bots digitally.

Alp Uguray (46:36)

So co -exist model with digital machines essentially, right? That's interesting. The second question that I have is what was one advice that stood out to you from a mentor that like it became the golden phrase.

Kaspar (46:41)

Yes. Yes.

One that we agreed with our co -founders was that in this startup, we will not do any mistakes.

Alp Uguray (47:07)

Yeah, it's something good to strive for. That's awesome.

Kaspar (47:10)

you

Alp Uguray (47:14)

And I think the last question that I have is, where do you think people are not building AI? Like where is it missing that you'd use for everyday life? Not negotiations or like chat GPT style, but more on something that's up.

Kaspar (47:35)

Well, I'm a huge fan of carments and nation states and one day most likely we'll go back to them because

It's so much more difficult to build innovation and technology in governments because of all the procurement reasons and because people don't want it because it's risk and you don't get votes if you do things people don't like. So no one does these things. So it's built into the governments that they are not progressive. and, but if you don't care about your votes or if you have some time to try things out and it's such a great playground for innovation because it first

of all it matters. The entire society is dependent on so many services from government and so AI definitely is part of that kind of how to engage with governments even in with ChachiPT as a government, how to detect viruses, how to do preventive medicine, how to do smart taxation, immigration, all of this kind of that today are handled to few people

with few advisors who build decisions based on their next elections is kind of a place where AI needs to find a place to disrupt the market.

Alp Uguray (48:58)

I really love that even in basic things like I want to renew my driver's license or apply for immigration or apply for, I don't know, like home buying process. Even those operational stuff, there's always a human agent somewhere sitting on top of.

Kaspar (49:16)

there, you need to come to Estonia because for private license, you get the notification that you need to renew them into a month. If you click there, then you need to verify that this picture is still correct. Their address is still correct. You sign digitally and then next day license is delivered to you. Yeah. So and all the other public services as well. So from taxation, one click.

Alp Uguray (49:36)

around.

Kaspar (49:44)

E -prescription, no clicks. So all of this is already so advanced that no humans are hired to do anything. But this is not even AI, this is just automation of government services and data exchange kind of. But AI building advanced smart layer on top of that now with that massive amount of data that it has access to. This is definitely to bet, but it's difficult. Difficult to make.

Alp Uguray (50:13)

And building that foundation layer is super important though, like having the data ready, having the processes ready. I feel like in the US, especially, it's as far behind when it comes to these things.

Kaspar (50:26)

Yeah, it's so scattered and distributed and humans have so many things that they're afraid of. And if you're afraid of, then you don't allow politicians to do that work for you. starts with, I don't know, a national security number or whatever. Like in Estonia, digital names are public. So my digital name is 387 -1201 -2796.

It's like Casper, but computers understand numbers better, right? So basically, and through that number, they can offer me services. They know who I am. I have digital identity, unique name, and then they can offer. And that's why I can do driver license, because driver license agency checked my data from other agency, checked my photograph by passport agency, checked my address in population registry. I gave the allowance to do that. And then basically I don't need to do anything. They just exchange that data.

And to do that, you need to use the name. It starts from that. And it's a long story and for another podcast, most likely. But this is where we need to start also in the US, if you actually want government services.

Alp Uguray (51:31)

That's a great point. curious for, I know I already said I'm going to ask the last question, so this will be the last question. So you built the e -residency program and it's being used by entrepreneurs all around the world. And with Pactum now, as you're already like still in Estonia, what some of the benefits

that the startup sees by being part of the residency program versus the ones that maybe some of the other incumbents are not be able to leverage because you guys have it.

Kaspar (52:15)

Yeah, mostly what eResonant just helps is it just helps to save you time to tackle with the bureaucracy. It helps to save costs when establishing a company. Just like 100 euros and then you can collaborate remotely with founders and employees.

no need to visit any tax authorities, anyone actually. And of course with the economic incentives of not being taxable at all, no income tax for global nomads and especially. So all of this, doesn't help you to get clients or revenue. It saves your time to do your business.

Alp Uguray (52:55)

you

Yeah, I think that would be the golden deal. Like the opposite side of you automatically get some revenue and customers. No, that's it. Thank you very much for the conversation today. I had a great time. I appreciate it.

Kaspar (53:15)

Thank you