In The Arena by TechArena - Uncovering the Future of AI Agents with Sema4

Episode Date: July 8, 2024

TechArena host Allyson Klein chats with Sema4.ai co-founder Antti Karjalainen about his vision for AI agents and how he sees these powerful tools surpassing even what current AI models deliver today.

Welcome to the Tech Arena, featuring authentic discussions between tech's leading innovators and our host, Alison Klein. Now, let's step into the arena. Welcome to the Tech Arena. My name is Alison Klein. Today, I am very excited to have Seme4AI on the show, and we've got the founder, Antti Karelainen. Welcome, my name is Antti Karjalanen and I'm one of the co-founders of Semaphore AI, a company we launched officially fairly recently. And we are working to bring intelligent AI agents to the enterprise, talked a lot about different topics about AI on the Tech Arena, but we haven't ever explored the topic of agents. Can you tell me a little bit about agents just to set the ground for the discussion today? Yeah, for sure. So agents are the most exciting use case for generative AI, kind of broadly speaking.

Agents will be technology that's going to drive some of the more exciting outcomes, I think. So what is an agent? It's a piece of software that contrasting to traditional software that we've been used to does not just make you more efficient at your work. Agents will actually complete the work for you. So you think about them as sort of a knowledge worker, assistant knowledge workers, or software that can reason, collaborate, and act with humans if you

wish. So a software that actually autonomously can complete work for you using generative AI. Interesting. Now tell me a little bit about how you started Semaphore AI. And I know that you've been involved in many technology startups in the past. Why Semaphore AI. And I know that you've been involved in many technology startups in the past. Why Semaphore and why now? Oh, good question. I think working on a concept like agents right now is probably the most exciting thing one can do. But my background starts with RPA, so robotic process automation. I've been involved in RPA for the past five years or so. Previously built a company called Robocorp to do that with an open source stack for RPA.

That was a really exciting journey. Personally, I started tracking Genitive AI a few years ago really closely, and it sort of naturally led into this world of agentic automation. And then I had the good fortune to be connected with the other members of the founding team that come from backgrounds like Cloudera, Hortonworks, AWS Docker, pretty seasoned and experienced team. And we got talking about doing a joint company around AI agents. Then we were able to work with some good investors at Benchmark

and Mayfield to make that a reality. But ultimately, what drove me personally here is, you know, the fact that I think this is going to be one of the biggest things that happen in enterprise automation for the next decades to come. Now, everybody's talking about co-pilots. We've talked about co-pilots on the tech arena. Can you just characterize the difference between applying a co-pilotots. We've talked about co-pilots on the tech arena. Can you just characterize the difference between applying a co-pilot to a problem and applying an agent and how companies should be

thinking about the two technologies? Co-pilots are pretty much everywhere right now. Every application you have is going to have a co-pilot and it's sort of maybe a chat window on the site that helps you use that application. Maybe it's something that generates text for you, summarizes things. So it's sort of an AI widget on the side, but never the main character, if you wish. Whereas the AI agent is something that you can actually interact with to get full work done. So instead of being just an assisting on the side, an AI agent is kind of the main thing.

AI is just going to be a new way to interact with enterprise applications, whereas a co-pilot is more of an additive layer on top. When you look at the solutions that you're targeting, who benefits primarily? And what progress have you made with customers? So we as Semaphore interact directly with line of business experts. We don't sell to, say, the IT team or the background teams. We interact directly with line of business and we actually allow the line of business to define their own agents with natural language. So right now we are focused on various internal operations teams,

whether that's finance operations or sales operations. The application areas are pretty broad, but to start off in the first customers, we are doing a lot of use cases around the finance teams operations and really allowing the finance line of business experts to define their own work in natural language and then turn that into agent. So that's really a unique approach where you don't have to be a software engineer or an AI expert to be able to design your own agents. Can you take us one click down into some of the top use cases that you think will be early deployments of agents in the enterprise? I think right now what we are seeing is a lot of the attention has gone into things like customer support because the benefits are really obvious. You have a lot of people contacting you at the same time and you can potentially shift some of those interactions directly to agents. There's of course some risk in that approach as well where the agent is interacting directly with your customers and so you need to have an escape hatch to bring that conversation back to humans.

Our approach has been instead of going outside in we have been going from inside out so finding these work streams internally that previously people couldn't address with things like RPA, things that deal with unstructured data, potentially semi-structured processes. So there we can make some new progress with agents and bringing them on. And so the use cases might be, if you think about internally, they might be things like HR onboarding, which is pretty clear to understand how an agent might be able to interact with your HR team to just ask questions and onboard a new employee in a matter of minutes rather than taking hours.

Now, I know that you are early in the stages of delivering this technology. Can you tell us about how you and your team have developed agents thus far and where you are with engaging customers? Yeah, we have first customers going through pilots and moving into production right now. Agents as a category of technology is pretty new. And we as AI industry collectively have just seen first useful agents appear, maybe it was end of last year, in areas like software development and assisting coding. But the way we are doing agents in the enterprises is still pretty new. And so these agents obviously need to be able to connect

to all the existing enterprise applications to drive automation into them and also enterprise data sources. So there's quite a bit of infrastructure to be built to enable an agent to be operationalized. And then, of course, we need to discuss security guardrails and compliance when you're dealing with enterprise as well. So building all of that just takes time. And when you look at how fast AI is advancing in the industry, I mean, I don't think anybody really saw this foment of technology coming out since 2022. And it's not like AI is new to the scene. It's just I think we've entered a different phase of its development.

How fast do you think agents are going to integrate into these enterprise applications? And what do you expect for the second half of this year? Well, I think that development is going to be incredibly fast. Being part of this wave, I think, is unlike anything I've seen before, being part of tech. And we're just seeing tremendous speed of companies executing from the fundamental technology of an LLM into building out the software frameworks that are needed to build an agent to the infrastructure that's needed to deploy an agent and operate. It has been maybe a year or 18 months going from really useful use cases of LLMs to having those LLMs fundamentally drive agents. So I think for the second half of

this year, we're going to see a lot of really good case studies, use cases come out and the first enterprises goal to scale out the agents. And the next year, I would think that it will be a year that we see agents go really mainstream. When you look at the market opportunity ahead, and we've talked about workflow integration, you talked about customer service. Obviously, I can imagine a ton of different workflows inside an enterprise that would be applicable to this. How far afield and how broad in terms of integration across the enterprise application suite do you see agents going? Is it something that you think will be pervasive in a few years, or are particular industries better suited for integration?

Or is it really a horizontal technology in your mind? It is really a horizontal technology. The way I think about value that's being created by agents is you can really think about three categories that we are seeing across our pilot customers. Which first is companies looking to save money, drive efficiency with agents. So this would be basically extending your RPA and business force automation program with agents. The second part would be avoiding costs. So better compliance, this would be applicable to banks and insurance, allowing, let's say, banks to better match with changing the regulation much faster

and avoiding potential large fines as a consequence of non-compliance. And then there's really ultimately creating new business opportunities, creating new revenue sources. So we're going to see agents go from cost efficiency, cost saving to new revenue opportunities as companies understand this more widely, but it's ultimately a horizontal technology and if you've been involved in RPA, you're seeing this revenue opportunities as companies understand this more widely. But it's ultimately a horizontal technology. And if you've been involved in RPA, you've seen these kind of very horizontal technologies being able to automate pretty much any application system out there. Agents is just

going to take that to the next level. Wonderful. Now, I guess my last question for you is where do folks find out more about the technology that you and the team are delivering and engage your team if they're interested in a trial or POC? So, of course, I recommend everyone to go to semaphore.ai. And we have a good series of blogs that go about explaining the basics of understanding enterprise AI agents. And then we are launching a first public product offering, which is Semaphore Desktop, that actually an application you can download and just go and build an agent on your own with it. So that's going to come later this summer. And there will be some courses that we'll publish around that. So Semaphore AI, the blog series is a good place to start with.

Well, Antti, thank you so much for being on the show today. It was wonderful to learn about AI agents, a little bit about Semaphore AI and about your background. So thank you so much for spending time with us. Thank you. Thanks for joining the Tech Arena. Subscribe and engage at our website, thetecharena.net. All content is copyright by The Tech Arena.