Jentic builds the missing layer between AI and enterprise systems: Sean Blanchfield, CEO at Jentic

0:00Sean Blanchfield:Hello and welcome to the Scaling Europe show presented by Deal. This episode is sponsored by SurrealDB, the multimodal database for AI agents. Omni, the AI analytics platform trusted by growing companies like Perplexity and Cibesia and VentureComet, the platform that gives startups and scale-ups real-time equity tracking, daily business insights and automated management information. Thank you for joining me. Please like, comment and subscribe.

0:33Sean Blanchfield:Hello and welcome back to Scaling Europe show. I'm Seb Johnson. Today I've got an amazing guest, Sean. Sean, thank you so much for joining me. How are you doing today? Yeah, I'm doing great. Thanks for having me. It's a pleasure. You're now building Gentic. For those who don't know, can you give a bit of context about what that is? Yeah. So in our view, there's a missing layer between the APIs and the AI when it comes to enterprise. I think everyone deploying agents at the moment or AI applications in general are realizing that the bottleneck isn't the quality of the models. It's not, you know, the path to AGI isn't what people are thinking about.

1:03It's how to actually hook it up to all of the different back ends, all of the API platforms, all the bespoke systems that exist in your company, especially if you're a large company, and how to do that securely and reliably. And it doesn't look like OpenFlow is the answer. So there's a new layer there, and that's where we sit. It's the enablement layer for AI in enterprises.

1:26Sean Blanchfield:And why isn't it something like Open Claw? You know, it's absolutely taken the world by storm. We're now seeing all these rapids trying to make it super accessible by everybody. But why is that not the solution? Yeah, well, yeah, Open Claw is a phenomenal thing to see. I do think this is a moment in history where it proves to us what's possible. But the risks are intolerable. The risks are pretty much intolerable even in your personal life, I think. And some people, you know, they're going to take those risks anyway just to be on the bleeding edge. if you have a small business, I would not hook this thing up to all of your bank accounts, your email and everything.

2:00If you have a business where there's a lot at stake, right? Or the surface of potential systems is very complicated. It just, yeah, it doesn't work from a security point of view. And it generally doesn't work if you just hook an LLM up to a thousand systems and tell it to not make a mistake, right?

2:19Sean Blanchfield:Yeah, that makes sense. It's one of the reasons why I've been very apprehensive to give it greater access than I have. I'm using it for very basic use cases. I've got it like a Telegram channel. But I don't want to give it everything because I just, who knows, feels a bit of a Pandora box situation. Yeah, it's a real tension between security and reliability, I think, like underneath the hood. If you look at Siri, you look at Google Assistant. I mean, in a way, OpenClaw really raises the question, why aren't they that capable? And the answer is because they're trying to, especially Siri, I think, trying not to screw things up.

2:54Right. So this is the real question. Yeah.

2:57Sean Blanchfield:And yeah, I guess that is the risk, isn't it? But does it also feel like a great moment in time for you and Genetic? Because all of a sudden, everybody's trying to connect to 100 APIs. Everyone's talking about OpenClaw and trying to connect systems in different places. And so is this like a good opportunity for you to step in and say, well, actually, if you are running a small business, if you are running a business where data security is really important, actually, you need to be thinking about this a bit more thoroughly. Yeah, well, we're very happy to see that the narrative is meeting us where we are.

3:25We're building from a pure conviction basis from first principles from two years ago, thinking, OK, we play this forward. These are where the problems are ultimately going to be in the integration layer. Security, discoverability, access control, old fashioned problems, really, but that need new fashion solutions. and we were out there alone for a long time and now there's a lot of people focusing on the same area, which is very validating and very good for us as well.

3:51Sean Blanchfield:And why was this the area that you decided to build in? You know, in your background, you built a lot of really interesting businesses. You're now building at this kind of API infrastructure layer for agents. Why was this the place where you decided that you could build a big business? Yeah, yeah. Lose a little bit of a connecting thread between that because the gaming stuff was the back end for Call of Duty, basically. and a hundred other video games, but we were the infrastructure provider, online infrastructure provider for video games. And that's where I really cut my teeth on this kind of big infrastructure stuff.

4:21And then as part of Activision, we were acquired by Activision in the end, seeing how you integrate with the broader CRM in a very large organization that might actually be hooked into individual player accounts and promotions and stores and all kinds of stuff. So it's exposure in that world that gave me the perspective to do this. but from an entrepreneur perspective obviously every entrepreneur has to start a business in this area as an entrepreneur I've learned sometimes the hard way the best thing is to do it in your wheelhouse where you actually have some leverage something you can really bring to the table where you have some kind of right to win and for me that's back in infrastructure distributed systems so this lines up with that the other part just speaking as a founder is trying to find somewhere where there's a platform to build that can have lasting value that isn't just going to get wiped out by Sam Altman someday and isn't just more naturally a feature in Copilot or some other Microsoft product.

5:17And you've got to go down the infrastructure layers to find that. And when I analyzed this, and I was quite analytical about it in the early days, I realized, and I convinced myself, there is a new infrastructure layer here. It's above APIs, but it's under the LLMs, and it's a real enablement layer, and the incumbents aren't well positioned to do it because a lot of these questions end up about sovereignty and stack neutrality and things like this. I believe, as you play this forward, if I'm running an enterprise, I need a neutral layer that sits above multiple different stacks and vendors where I can develop my workflows to govern my workflows, observe my agents, and manage my automated business.

5:57And I can't just buy into any one ecosystem for that, which leaves the space for a new layer of companies to exist. and that's where we are aiming, Gentic.

6:06Sean Blanchfield:Yeah, and I guess it kind of rings true throughout all of your career. You know, the founder journey that you've been on, although it's been quite diverse, has always been a theme of working infrastructure layer, whether that's in gaming or now AI world, you've always been doing infrastructure, I guess. And what kind of advantages do you think that's given you over other founders? Advantages and disadvantages, maybe. but uh what's what i find is is different about us uh for better or for worse is that we are really just moving forward on a first principles basis you have this technology we need to connect ai into the real data of the world valuable data of the world right and this isn't you know slack and notion this is a big enterprise systems where something like 80 of the software in the world is an enterprise and all the like valuable data you know really valuable data bank accounts and stocks and ERP systems and logistics and medical records and everything.

7:02It's that kind of stuff. So how you connect that in that, you know, when we were developing this business and trying to figure out which way it's going to move and what's going to be necessary, people were barely talking about agents. No one was talking, MCP didn't exist, so no one was talking about integration. No one was thinking about, or very few people were very clear anyway about security and access control and observability and the need for a unified layer across these very very complex api states that you find in large companies and i mean i mean i guess a lot of your your listeners will will know this firsthand but for those who don't like enterprises will have thousands of apis that they know of sometimes tens of thousands of apis that they know of and that's how that business runs so for them they need to hook that up uh so what maybe sets us apart is just trying to approach this basically just from first principles what's required and then just making that bet and going forward saying we're building that it's going to take some time but you know everyone else will meet us there because this is inevitable um and when you go moving from a more product-led perspective and i've done that in previous companies you know the whole lean methodology and all this stuff it can lead you into what in computer science we call hill walking where you're just doing the next obvious thing you can get to local maxima you know this is a good solution to this particular problem but you can't really make big leaps you know big leaps of vision and say, this is where it's going to end up and this is the real big opportunity.

8:27And we just need to keep our eye on that prize and just build towards it. And that is the kind of business we're doing now. This is a bet on building a new kind of platform for enterprise. Not a feature, not a point solution.

8:42Sean Blanchfield:And what do enterprises need to do to get themselves ready for this new world of AI? I mean, you're building this infrastructure to connect to APIs. Are there enterprises that you're working with that are ready to connect to your platform and get the most out of it? and what makes him special? I think where they are mostly right now is 18 months into pilots and realize when it works, it works. But how are you going to have a hundred of those things or a thousand of those things? How are you going to manage that in the future? How are you going to build them and hook them up and be able to stand over it from any security or compliance perspective or just understand how your business is running and what you should do next?

9:21So basically scaling problems. So where we're going in is really, it starts with API enablement. An enterprise who's been keeping their own technology stack in good shape for the last 10, 15 years will have an API platform internally. Some don't, most do. The quality of that has been for human developers. First step is to lift the quality of that. And we have AI to help basically help document those APIs and make them better. And then you have to put agent experience on top. Can an agent connect in and find the right API in the moment when there's 16 ,000 choices? And when it does find the right one, can it use it well?

10:00Is it documented to the extent that an LLM in a single shot will get it right? And can it call it security? It's securely without the LLM being responsible for calling it securely. You don't want that credential and that access control living in an LLM prompt. You need it in a centralized system. So that's like plumbing layer stuff. And that's the entry point where it is today. and a lot of enterprises are there and broadly that's called AI enablement or platform AI enablement by these enterprises today and that's working very well. And we basically tooled up through that layer to use LLMs to help solve these problems that have traditionally been very hard, like integration.

10:38And all on open standards, by the way. If you're running an enterprise right now, you have different paths ahead of you and I think that a whole bunch of those paths leads to being dependent on a whole bunch of wall gardens and vendors that are new and mightn't be around for a long time. And there's another path which is about autonomy and open standards. We're making a bet on that second path. Yeah, and then you get into workflows. Workflows bring you into governance. And then, yeah, that's basically the rough shape of the business.

11:11Sean Blanchfield:Are you seeing similarities between those that are able to move the quickest and do it the best compared to those who are maybe not getting the value out of those pilots? Yeah, I think it's got to do with foundations again. So a lot of businesses have had to push through a lot of pilots in the last while. And incidentally, some things that get called agents are, you know, I'm a bit skeptical of, right? So yeah, I'm sure you've seen a lot of things called agents like me. There's a script called an agent, an NNN or a Zapier workflow called an agent, OCR called an agent. so there's lots of that going around or just just chatbots that are called agents or copilets but by by uh in my book we these things if you're going to use a new word it should have a new meaning and an agent is a an llm driven process it chooses what the next step is um so to get back to your question a lot of enterprises have tried a lot of things and some of them have worked and some of them haven't and the the ones that are struggling have just done point solutions which makes sense they need to show the business value and sometimes they do but now they're now they're tripping up under the weight of all of that the technical debt is sometimes overwhelming they might have tons of n and n workflows that uh i mean kind of hang together but maybe have a whole bunch of embedded code and you know it's not even very well written it's actually sometimes worse than software.

12:36You know, if it was actual software, it would be under software management somewhere in the company, run by people and maintained by people who know what they do. But if it's in any of the things, it's just better JavaScript code where no one can look at it and paste it into a text field. So that's roughly what bad looks like right now. And the critical question for those guys is how to get that organized so they can move forward. How to put in place some kind of unified platform-based approach so that they can have lots of what they've proven out in isolated cases. And good right now looks like that.

13:06And in the best cases we've seen, whether it's people we're working with or companies who basically roll their own because they can't wait around for a vendor like us to emerge as a leader, it's getting really good API management in place. These are the fundamental tools, or at least building blocks of the tools that the AI needs to connect to. And then putting in centralized discovery and security. And we saw that, I believe I saw Deutsche Bank, Like for example, giving a presentation, describing that. And once you have that, you have something you can hook agents up to without having a heart attack about what they're going to do next.

13:41Then you can talk about workflows, right? Which is putting reliability back into the equation. That it's like a high level tool, right? If something can be done deterministically, it really should be done deterministically. It's optimal to do that. You capture that logic, you put it in a workflow, regardless of the workflow orchestration platform. There's so many right now. And our bet is actually on open source and open standards. Eradso is a new standard for this. I think that's very important for all kinds of sovereignty reasons. And when you have workflows, then now you're getting into a really high-performing system.

14:16You have LLMs maybe in production, calling workflows that can reliably and repeatedly and efficiently, by the way, with low energy, perform useful work. another thing we're seeing is this the phase run right now I'm sure you've seen it like it's workflows everywhere whether that's iPaaS platforms like Zapier whether it's new entrants like N8N whether it's like Gmail building and workflows and agent kit and chativity everything has workflows now this is all about workflow orchestration we think that's the easy bit like running a workflow. You've got some JSON file listing a bunch of steps and just running each of those steps.

15:01We've in fact, we've open sourced a library that'll do that for Erato for the open standard. And we're about to open source an editor for it as well. The question beyond that is where do all the workflows come from at scale? If we refocus, what's the goal of doing AI as an enterprise as a business? I would argue it's automation of the things we couldn't automate before at a scale that we couldn't imagine before. So if we remember that that's the goal, it's not just to do lots of activity and say we have lots of AI initiatives. That automation takes the form of workflows. The question is, how are you going to create 10 ,000 of those and more importantly, maintain 10 ,000 of those in the face of constantly changing business requirements?

15:44The answer is agents are obviously going to be creating the workflows. This is, I think, the next stage of the conversation. I would expect to see emerge this year that our attention moves past how will I run workflows because that's a dime a dozen, right? And there's open source alternatives into how am I going to create and agentically manage and maintain workflows at a scale that humans could never do. And that's the real game. And the answer has to do with sandboxes and testing environments and that's where we're focusing a lot of energy.

16:15Sean Blanchfield:Yeah, I want to touch on something you said about open source and open standards and that being the big bet that you're making. what does the world look like if you're wrong about open source and open standards i think in the worst case scenario it's still a really big slice right um in the alternative scenario it's about do like leading vendors uh obviously all the hyperscalers have their own solutions here um and or or will soon for workflows um and then the large ai companies have their own, right? And in ChatGPT, we already see it. So the alternative would be that companies have to pick one.

16:55I think in any company, you end up wanting to pick one, right? To have workflows, it's, I mean, you can pick 10, and by default, you'll sleepwalk into having 10 in your company. But then your next critical thing will be, we have to get this under control. I mean, at least we need to be able to answer how many workflows we have or how we're doing. So they'll try and centralize, and they'll try and pick one. So in the alternative, some companies will say we're going to be an Azure house and just continue to do it all on a Microsoft stack. And that's fine. And many organizations are like that. But the wind is roughly blowing in a different direction and has been even before AI, which is towards multi-cloud and cloud neutrality for good reasons.

17:35And I think those reasons are made even more acute now with AI when what's at stake is your very business processes, like how you run your business. we're encoding that into workflows and handing it over to machines to run. So in the future where you're just keeping going down that path and you're, especially if you're a services company, literally all you have is a bank account and business operations, right? And all those business operations just became workflows. You have a choice. Do you want to put all those workflows on one vendor? And then basically what you do with your business is entirely dependent on what that vendor allows you to do.

18:13and the rent that you pay them is out of your control. And whether you want to take your business and say, you know, actually, I think we need to take these processes and move into a new market. We need to adapt them. And whether you can or not will depend on that vendor. The other path is you go on open standards, right? And that means that those workflows belong to you. They're actually, you can look at them as IP. They're private to you as well. They're what makes you different from your competitor. and you might consider them core business assets, core business IP, that you have full autonomy over and no one can take it away from you.

18:47Sean Blanchfield:No, I agree. And it definitely does feel like the way that the wind is blowing and especially with the key points around data sovereignty and strategic autonomy. But I also want to touch on Ireland. You know, you're building from Ireland, from Dublin. I saw a Microsoft report that stated that AI adoption in Ireland is one of the highest in the world. What's it like building that? Is there a culture of openness and adoption to AI, or is there scepticism? It's fun these days. It's changed a lot in my career, which is like 25 years or a little bit more of doing this kind of stuff. Like when we were starting out, the tech industry in Ireland was not so developed.

19:25It was a lot easier to hire very smart people back then as well. And since then, as so many US companies have set up in Ireland, we've been obviously very successful positioning ourselves as a highly open economy that's friendly in all kinds of regulatory and tax ways. That's the hand that Ireland Inc. has played very successfully. And now that talent is incubated in what's really a very small society, it's incubated long-term. And I think that's where that openness to technology, that's that about people using AI. I think it's probably just directly related to so many people in Ireland working directly in big tech and telling their friends.

20:04you know it's it's so much of the economy the economy is based on it um so it's very tech forward from that perspective so it's fun um obviously like every country we have all kinds of challenges and when it comes to ai like i'm on the uh ai advisory council to the government here in ireland and we think i think we think about this stuff a lot uh all the pros and cons and challenges and opportunities um and maybe one of my main hobby horses is thinking about the economy of the future and whether you actually get to be in the supply chain of AI when everything gets automated. What does business of the future look like?

20:39Is the wealth distribution or is it just like accelerated aggregation of capital towards business owners, in particular the business owners who own the AI, right? And as a small economy, where do we contribute? I think that's like a critical question, maybe for all of Europe, to be honest. For Ireland, it's not realistic for us to say we're going to build the next breakthrough LLM. It's not going to happen here. But there are other things we can do. And some of that is maybe fix energy, if we can actually fix our grid, like the demands have grown too fast and the grid's falling over. We can build data centers.

21:16We host an outside amount of the cloud because of the same companies with so much like, you know, the EOS data centers and Amazon are all here and so on. The other part, maybe regulatory, might be a nice regulatory base, as we've been for GDPR, at least from the company's perspective. But I think actually more legitimately, the opportunity to take AI and apply it in large enterprise, because we have these natural clusters here because of the last 20, 30 years of direct US investment now with real operations. like and you know Dublin's like the size of Manchester or something we're a small relatively small city uh you can walk across it in 20 minutes and people know each other and we in terms of being able to go into like large global financial services companies or large farmer companies large tech and so on um we we can and insurance right we can actually do it in these natural clusters that are right here so that's what we're attempting to do here in Gentic and it's actually quite fun now.

22:19It's not just the camp management and outsourced jobs here anymore, but there's C-level people in global companies in the small city and we can get in and it's a wonderful beachhead market for proving out the technology and how it gets applied.

22:32Sean Blanchfield:Yeah, I mean, Dublin's an amazing city and I guess it benefits from being small in the sense that you get this real great density of talent a bit similar to somewhere like Stockholm, I guess. But look, Sean, thank you so much for joining me. I'm a huge fan of what you're building. Let's stay in touch. yeah that was wonderful thanks Ed I enjoyed it