Digital Workers, Human Work: Darko Matovski on Agentic AI at Scale

Darko Matovski:

In the previous wave automation, people try to do what the humans is kind of like trying to build a plane by observing birds.

I think now we figured out to really accelerate, we need to reimagine and you know, we need, jet engines and so on, but reimagining the whole workflow and the first causality there was kind of software written for humans. And now we're seeing enabling people to do more meaningful jobs, going away from the repetitive, the mundane, which no one really gets excited and people, when they're not excited, they make mistakes and so on, to being able to have a very clear delineation between what is a human work versus what is a digital worker work.

Jeff Kavanaugh:

Welcome to a series of conversations with leaders who are turning AI's potential into measurable business results. We're here at the Semafor World Economy Summit. I'm Jeff Kavanaugh. Today we're here with Darko Matovski, Co-founder and Chief Executive Officer of causaLens. Darko, welcome.

Darko Matovski:

Thank you very much for having me.

Jeff Kavanaugh:

As a leader of a company that has digital workers, what is the most surprising impact of the AI revolution and how will it redefine how companies compete?

Darko Matovski:

We are seeing complete rewriting of how companies operate. And we today see it only with the most forward-looking organizations, but this is going to come and become mainstream.

The first thing we observed is that we no longer automating exactly the way people did something. We are taking a step back and we're reimagining the process. We're saying instead of five teams and three software products needed to achieve a job, could we reimagine this and have a digital worker that will do this end-to-end?

So the first thing that we observed was that software was built for humans. Once you have a digital worker, you no longer need software in the current form with like UIs and buttons and of course, digital worker is a software, but you don't need the human software.

In the previous wave automation, people try to do what the humans is kind of like trying to build a plane by observing birds. I think now we figured out to really accelerate, we need to reimagine and you know, we need, jet engines and so on, but reimagining the whole workflow and the first causality there was kind of software written for humans. And now we're seeing enabling people to do more meaningful jobs, going away from the repetitive, the mundane, which no one really gets excited and people, when they're not excited, they make mistakes and so on, to being able to have a very clear delineation between what is a human work versus what is a digital worker work.

In previous ways of automation, we've had those two things being intertwined. Whereas now we're seeing a very, very clear separation of where does the digital worker stop? Where does the human start? And we're seeing a lot more, humans being more humans. We spend more time with other humans, thinking about big ideas, new business models, sales, all these things that require trust. Whereas the digital workers are doing all the mundane, boring stuff.

Our vision is that, and we hope that at the end we'll end up with humans spending less than 14, 15 hours a day staring at a computer screen and spending more time with humans.

Jeff Kavanaugh:

This uncertainty requires frequent adaption. So you're talking about keeping things consistent. How can processes though adapt to the outside world, like a digital worker, are they able to also adapt and change?

Darko Matovski:

A lot of people currently talk about calling it the build phase. And it's normal because we don't have a lot of digital workers. So there's a lot of obsession around how do we build them. And we have a really cool product there but that automates it.

But the more interesting part is how do you enable, how do you create an operating system that can keep these digital workers alive, learning from humans, learning from their own selves, learning from interacting with other systems in the environment, learning from the internet and create a continuously self-evolving system that doesn't require going back to the IT team to rebuild it when the world around them is changing?

I think that is the key for scale and we spend most of our time thinking about, how do we enable, how do we create the platform that enables that vision?

Jeff Kavanaugh:

With all this investment out there, not just data centers we have to build and everything else, where are you seeing the measurable ROI?

Darko Matovski:

The way we approach it is we need to know what is the estimated ROI before we do something. So we have a transformation consultants that will look at a workflow that was previously done by humans or outsourced or with software. We estimate the annual cost of doing that work. And we typically look for at least half a million of annual costs. And then we say, okay, well, if we can, if we are going to tackle this, we need to deliver this at a, you know, a fifth, let's say, of the cost.

But now it's kind of predictable, we know the amount of investment required to get that ROI and we can now also tackle it very quickly because we know what the process or the workflow was before. We know what needs to be done.

Contrary to popular believe the inference compute required to run this sort of digital worker is tiny compared to the economic benefit they're creating, right? Because you're comparing, you know, token costs with kind of human costs. And those are kind of orders of magnitude difference. So if you're very deliberate and you imagine and reimagine a process in your, you can calculate the potential ROI and only go after high value use cases, you end up with like fantastic ROI. There's a lot to go around.

Jeff Kavanaugh:

You talked about doing this based upon estimated savings. You wait until people actually realize savings. So that requires a lot of confidence in your models. How did those evolve and how is that being received in the market?

Darko Matovski:

You start paying for the digital workers when they're built. So we take a lot of risk in the training phase to make them suitable for your environment. And as you say, like it does require quite a lot of confidence to do that. And we didn't start there, but now that we have repeatability, we've gone and done that multiple times. We know what is likely to work out and what is not. And so we only engage in building digital workers that we know will deliver value.

In fact, we go beyond that, we have pre-skilled digital workers for certain functions. All of the back office, finance, accounting, procurement, HR. We've already pre-built these things. Also, we have very specific domain, pre-skilled digital worker in finance and life sciences. And so we kind of invert the problem. We say, hey, for you two, and we now have an entire operating model around this. We can come in and very quickly reimagine all of your back office, for example, or we can come and reimagine all of your commercial pharma use cases, because we already have these pre-skilled digital workers. We just need to switch them on and calibrate them in your environment.

Jeff Kavanaugh:

When the agents are working, who do they report to in the company?

Darko Matovski:

So the vision here is that, and one of our customers, I think summarized this nicely. They're like, we want to get to a point where one, you know, flesh and blood human has at least like 100 digital workers that work for them. And I think that's the end state. We want everybody to be able to create their digital employees and manage their digital employees.

So we are building this capability on our platform, which we call the Chief of Staff, where it allows you to observe your digital employees, your digital workers, and train them, coach them, switch them off, fire them if they're underperforming, promote them to do bigger things when they do well. And I think that's where we're going to end up. Everybody becomes a leader.

And I think that's going to be pretty exciting future where jobs become more enjoyable. We all have to do a lot of boring things in our daily life. Now imagine just doing the fun bit and just doubling down on that.

Jeff Kavanaugh:

What do you think the ratio is going to be if 100 is the number of workers today? You said in our earlier discussion how they're decoupling now. You're not simply just having digital and humans kind of doing the same things and one's taking over the other. They're literally delineating into clearly different roles. What's the ratio of people relative to that 100 that we'll have in the future without it's growing and things like that?

Darko Matovski:

Yeah, I think that's hard to answer. It will depend a little bit on the business. I think we're going to, over the next few years, new businesses are going to be born, new roles are going to be created.

I think, yeah, I just remember a few years ago that the job data scientist was like the hottest thing ever, but 10 years before that it didn't exist. So I think we're going to get a bunch of new jobs that don't exist today. I think humans overall are going to end up doing more human things.

Jeff Kavanaugh:

Yeah, I also heard yesterday in one of the sessions how, let's say of a hundred projects you want to do, people are getting to eight major ones because of capital today. And let's say you're 3x more productive. So now maybe you'll get to 24 of the hundred. So it's a long pipeline of things you want to do, customers you want to serve, products you want to develop. So maybe it's, that's the cup is half full approach anyway.

Darko Matovski:

No, for sure. And I'm bullish on humans. I think we are very adaptable, we're very creative, and I think we all have a big backlog of things we always wanted to do, and now we actually get to do them. I feel it's like AI is this liberating technology that liberates you from the grunt work and enables you to do all the things that you always wanted to do, and I hope that's how it's going to help the humanity.

Jeff Kavanaugh:

Darko, thank you for your time.

Darko Matovski:

You're welcome. Thank you for inviting me.

Jeff Kavanaugh:

I'm Jeff Kavanaugh for the Infosys Knowledge Institute. Until next time, keep learning and keep sharing.