AI Adoption Speed – Techstrong AI Podcast EP33

Hello, and welcome to the Techstrong AI Podcast. I'm Amanda Ani, and with me today is Magnus Rein. He's the CPO of Open Stream ai.

How are you? I'm doing fine, and thanks for having me, Amanda. I'm glad to have you on the show again.

We spoke about a year ago. Absolutely. Uh, in the AI world, that's, uh, dog years, right?

So it's seven years ago, um, it feels like, Yes, it does. AI has been advancing very rapidly, but we're here to talk about why it is that AI adoption seems kind of slow. So let's start with, first of all, tell me a little bit about Open Stream so our audience knows about Open Stream, and then we'll go into that topic.

Of course. Yeah. Open Stream is, um, AI provider, the provide new conversational ai, which is the automation of conversations for customer service and for employee engagement.

Um, we also do, you know, a lot around language AI from large language models to symbolic systems, uh, and agent packaging of those. So, you know, broad set of capabilities in our platform. So that's, uh, that's what we do.

Okay, wonderful. So then let's get to it. Why, with as many advancements as AI has shown over the last year or year and a half, why does it still seem to be a little bit slow in its adoption?

I think it's nothing surprising if, you know, these are cycles. Um, there's nothing surprising in it. If you look at other tech booms going back, right?

Um, I, I've started my career in the internet, boom. Uh, and you know, e-commerce came in 1994 was some of the first e-commerce sites, uh, on the internet today in 2014, 30 days later, uh, 30 years later, sorry. Um, e-commerce is still not larger than physical retail, right?

Uh, so even though it's amazing to be able to just click, you know, deliver with Amazon Prime next day, stuff like that, it still hasn't dominated. The adoption curve still is, you know, way over 30 years. Uh, and you see that all kinds of technology, right?

Uh, capabilities come really fast, and then we struggle for 10 years to figure out how to package it to solve actual problems. Um, and, and that's the, you know, uh, it's, it's a long tail, I think in where technology is easy to advance, but human psychology and government regulation and stuff like that moves very, very slowly. So, you know, when you put those two things up against each other, it's, I'm, I'm not surprised, right?

Uh, it's, uh, the cycle of technology and we, we tend to be way too optimistic on technologies capabilities, and we overvalue, it's in, we over rate this impact short term. Uh, so we think like, oh, if AI can do this, it's gonna revolutionize the world in a year, but we also tend to underestimate the impact something has in 10 years, right? Uh, so the second order consequences, stuff like that, we can't keep track of.

So, so we kind of underestimate that long-term tenure impact, and people figure out how to package it to solve actual problems, right? Um, so, so, you know, uh, I'm not surprised, but then of course I would like, I would like everything to move faster because it's, uh, it's engaging and fun and I'm in the middle of it, right? But, uh, um, I think like, I think, like you said, there are a lot of business leaders and companies that are jumping on AI trying to harness it, but without any, uh, true problem that they're trying to solve without a, a clear understanding of what they really wanna address with ai.

So that, that's part of the problem. And, and then like you said, putting too much faith in AI when it's really just one additional tool, it's not the end all for Yeah. And, and it's hard work.

That's the, that's the second thing, right? That, uh, one thing is what it can do out the box, which, you know, on, on the surface seems, uh, um, amazing, uh, in, in modern, modern models. Uh, but then you figure out sort of, okay, how can I get it to consistently, consistently deliver value inside a specific problem, right?

And you find out that you have to build a lot around it and experiment a lot and test a lot and, and, and things like that to get the kind of consistency you would get from a human doing the task. Um, and, uh, that is a lot of hard work. Uh, I I used to say that automation is not automatic, right?

It's, uh, we think that this, this technology is amazing, so it can just work outta the box, but, you know, it doesn't, Yes. And I think part of it too is they're not quite seeing that return on investment. They thought AI would save money, but that's not quite always the case.

No. Um, it, it's partly also down to what, what you measure and what you don't measure, right? Um, so, uh, so take an example, right?

If you have, um, a system using ai, uh, to help you do something like, like, um, uh, uh, evaluated accident reports, uh, to, to see which ones are suspected works, right? So how do you measure kind of the RI, uh, from that? Uh, it's, it's not straightforward because you're already, you know, looking at all those and you're picking out and you have some heuristics to what you pick out and stuff like that.

So you have a c semi-automatic system in, in place. Uh, so you have to measure, and, and, and it might be because you are not settling these things quickly, it might take two years down the line until you have the numbers to actually check if it actually improved it, right? Um, and then you don't consider a, at any point in that, you know, example, uh, you might not consider the time from that, you're able to cut down the process with, so a process that took two months and people got, you know, the feedback two months after an accident.

If you get down that down to a day, the knock on effect is customer service improvements, uh, that you are not measuring. Uh, and people are super happy because they get immediate answers, but you're not measuring that. You're measuring kind of the, the savings.

So, so a lot of this goes down to, uh, not measuring the right things. Um, and in, in a lot of cases, uh, uh, to be fair is that, you know, uh, the, the, the crop of leadership in in, in it, right? If you look at CIOs and stuff like that, and, and especially big companies like I work with, they're, they're super savvy.

They know that the, the, the projects that they're engaging in, they have a, a, you know, targets for what it'll bring them, but they're also looking at the secondary effects, which is what are we learning as a, as a, as a, as a company, what are we learning about this technology so that we, you know, when when things becomes more mature and we're buying more systems and we're doing more projects, uh, we're, we're starting from a point of knowledge instead of starting from the point of we have to, we have to make it right the first time, right? So I see a lot of experiments and proof to concept stuff like that for, uh, you know, partly for the purpose of learning, but also, you know, for, you know, they scale them down a little bit, um, to, to get, get value. Uh, I do not see, except in, in venture capitalism, I don't see, um, the kind of projects I saw the last AI boom around 2014, right?

Uh, because then you had this AI boom, and, and all you saw was like a hundred, $150 million projects. Uh, and, and it was just, you know, um, total failures. You saw all the handshakes in the media, but you never saw the delivery of it, right?

Um, so, so, so you kind of had that, um, and this time you see a lot smaller projects. So you can see $10 million projects instead of $100 million project. Uh, I think, you know, maybe the venture capitalist, um, space that is investing billions and billions in, you know, um, wrappers around shat, who should, should perhaps, uh, look at what the, the big companies actually doing and spending, What do you think is needed in order for more, uh, AI adoption?

What, what, what does, what needs to happen? First off, it, it understand, right? Uh, it is a technology that's amazing and, and disappointing depending on how much you understand, right?

So, so if you understand the limits and how it works, you can do amazing things you couldn't do two years ago, right? Um, but if you don't understand it, you will apply to the wrong things, right? And you will, you will encounter, say, you know, failures that you don't comprehend, this should work.

You know, why is this working? Because you don't understand technology, right? So, and that this is also why adoption always takes a little bit of time, right?

Because understanding has to sit through the, the ecosystem so people start to understand what this technology is. And, and understanding is kind of, we're building on the knowledge of others all the time, right? It, it's not like, you know, the isolated genius is, is a myth.

We're always building on somebody else. So, um, so, so for this to kind of, uh, uh, the terms we're using to describe things, the, the, the, the metaphors, the reference frames, the, the, the projects and the products we we're highlighting and the methods that people are using to, to get value from it all, that has to kind of be assimilated and also, you know, produced. So you have this kind of gradual kind of building of understanding that is, that has to happen.

Um, and, and until, you know, the, the understanding is, uh, is better, uh, people will apply to the wrong things, right? Uh, and they will try to do, uh, do things that it's, it's a very poor fit for, uh, for the technology. Um, and just, that's just the way it is.

And you, you can argue that if people don't do those failures, you know, the, the market might not learn. So we had to kind of think that, well, it's a necessary with, with all transformational technology, you know, uh, people are talking about the bubble and stuff like that. It might be that that bubble is necessary to learn the limits of the technology.

And, you know, uh, it, it might be sort of like, like, like, like, uh, if, if, if you can believe a bubble theory, right? That, that you will have a kind of like a, a crash of evaluations and stuff like that, but 10 years from now, it might be seen exactly as well. It was something that was needed to get acceleration and, and learning in a short amount of time so that we could get that value 10 years down the road.

Yes. And I think, I think probably many business leaders are intentionally, purposefully holding back and watching what other companies are doing, what works and what fails, and seeing what do people really want with AI before they jump in. Absolutely.

Uh, I see some of that, but I also see that, uh, there's lots of projects out there that are internal, right? Especially when we talk about ge uh, generative ai, uh, in particular, um, there's loads and loads of use cases in empowering employees to do better quality work or stuff like that. And companies might not talk too highly of it because they see that, well, it's a one year project, it's one half year project, it's like half year project.

Uh, if we can get that under the radar and get the 20% productivity increase among highly skilled, highly expensive workers, uh, it's competitive advantage. Uh, yeah. So, so, so, so you, I see a lot of those projects, right?

Um, that, um, that, that happened quietly. Um, and, and I think that's, uh, that's a savvy thing to do, right? Uh, but usually those projects are high up the organization because, uh, you know, there's no history to point to this company over there, did that so we can get the same savings, you know, it's none of that happening, right?

So, so it has to be based on a belief, a bet, uh, uh, you know, something. And, and usually that requires you to be high up the organization to be able to say, we're gonna, we are gonna spend 5 million on this, and, you know, these are the benefits we hoped to get from it. Um, mm-Hmm.

So, so yeah, those projects exist and I think a lot more than people think. Well, if there was one key takeaway that you could leave our audience with today, what would that be? It would be that we're not over the experimentation phase, right?

Um, I, I, uh, I, I talk a lot about ai, uh, on, you know, what it is, what it can do, what it can't do, you know, try to kind of elevate understanding of, of people. And I still find that, you know, uh, naturally, uh, the, the, the majority of the world doesn't really understand artificial intelligence, uh, uh, uh, intelligence, uh, for now, right? And so I think the, the, the major, major thing that, uh, business leaders and, um, people that are not data scientists or, or, you know, work in the AI industry can do, um, is to learn as much as possible, uh, to learn through doing, not trust salespeople talking about, uh, you know, uh, the technology and claiming, you know, artificial general intelligence in the air and stuff like that.

Uh, uh, you know, everything has to be sort of tempered. Um, no, the intention of, uh, you know, why things are said in, in the AI space, there's a battle for legislation. There's a battle for copyright, there's a battle for, uh, uh, for market share.

Uh, uh, so, so, so there's these battles going on, which is naturally, uh, you know, shaping the narrative, uh, of some of the public, uh, messages that, that go out there. Uh, and we have to be kind of skeptical. And the only way to evaluate that, you know, what, what people say about artificial intelligence is for, to have an understanding of it, uh, to have tried it, to have worked with it, to have tried to apply it, and failed to have tried to apply it and succeed, right?

Uh, so experiment is much impossible. That's, that's, and that's a fun thing to do. Uh, you should negotiate to get, get time to do it right?

Uh, so, yep. Sounds good. Well, thank you so much for coming on the show and sharing your insights with us today.

Thank you for having me. All right. And thank you to our audience as well, and stay tuned because.