The AI Tax: Why Your Agents Cost More Than Your People and What That Means for Scale

00:00:11:02 - 00:00:32:03 Michael Welcome to disambiguation. I'm your host, Michael Fauscette. Each week, we interview experts in artificial intelligence, generative AI, and business automation to help business leaders understand how to use these tools for the biggest business impact. 00:00:32:06 - 00:00:42:06 Michael Our show today is the AI.

Why? While your agents cost more than your people and what that means for scale. I'm joined by Joshua Gould, CEO of The Big Word. Joshua, welcome.

00:00:42:08 - 00:00:45:15 Joshua Thank you very much for having me, Michael. 00:00:45:18 - 00:01:11:25 Michael You know, I, I have very interesting background, and I know you've been involved in language services for over 20 years. Co-Founder TBB global. And now you're running one of the largest language service providers in the world, 80 countries, 250 plus languages.

And you've certainly been living through the AI disruption for a lot longer than most of us, right, since the late 1990s. 00:01:11:26 - 00:01:21:28 Michael So I'm curious about what you know. Tell us a little bit about yourself and then about the big word and a bit about your journey to AI today. 00:01:22:01 - 00:01:43:25 Joshua Okay.

Well, originally I'm from the UK. I was born in the north of England and I went to college, and after three weeks of sitting in the lecture theaters, I thought, oh my God, you know, I'm way to ADHD to do this for the next three years. Plus, I want to go and actually make some money rather than make debt. 00:01:43:25 - 00:02:16:15 Joshua So I went to work in actually for cause brewers and you know, cause is not really known in the UK but it's brands are and I learned to sell and learning to sell was a really good kind of university of life.

And then I took a job selling language services to the US in the night, because I live with students in a student house, and I wanted to be able to sleep in the day just like them, and then work at night. 00:02:16:15 - 00:02:43:20 Joshua And then after the call center closed, which was roughly ten, 11:00, because we were selling to the East Coast, I wanted to go in party, and it was a great life for a little bit. I remember my rent was $60 a week for a room and, you know, so I could do that.

But at some point I realized if I want to, you know, really do something great, I had to be where the action was happening. 00:02:43:20 - 00:03:06:02 Joshua And that was on Wall Street, because I was selling language services to banks. And so that's how I ended up in the US. So about 21 years old, half a lifetime ago for me, I moved to the US.

And this is the official story. The unofficial story is the parties look great. I'd watch sex and the city and I thought, I want a piece of that. 00:03:06:04 - 00:03:40:21 Joshua Why would you?

Yeah, exactly. Or maybe I got the unofficial and official stories mixed up, but either way, that's. Those were the drivers. And back in those days, we had had major disruptions because.

And I'll intertwine the two questions here that you asked. We had just overcome a major AI disruption. We didn't call it AI because and I think it still is called amongst the tech community, machine learning. 00:03:40:24 - 00:04:24:20 Joshua I know AI is is sexy, but we had a machine learning disruption.

And really what it was was the reuse of previously generated content and where they could check it for in context matching. And you could run statistical likelihood studies against segments. So we could look at a document and figure out whether a sentence or which is roughly a segment was a fuzzy match or an exact match or an in context exact match to something we had previously translated, and then we could send this document segmented in a simple. 00:04:24:22 - 00:04:50:28 Joshua We used to call them GUIs, but now I think user interfaces.

We dropped the G and I'm aging myself here for your audience. And we create symbol interface. And then the the translators. Good.

Give us a 30% discount on things that were what we would call a fuzzy match. If it was exact, it would be a 90% discount. 00:04:50:28 - 00:05:22:28 Joshua And if it was an in context exact, it wouldn't even be translated. And that really was a huge disruption, actually meant that about 70% of our revenue would disappear overnight.

So we had to then apply wraparound services. And this might sound familiar to today to the AI. And we decided the best course was to offer integrations into the back end of these banks, content management systems and authoring tools. 00:05:23:01 - 00:05:48:26 Joshua And I'm talking about when a content management system wasn't running just a website, you could actually send it to your printers.

And so we went to the big banks HSBC, UBS, Barclays, big banks that were in lots of equity research, and they were then trading on this intelligence that was publicly available, but not necessarily in the language their traders were speaking. 00:05:49:02 - 00:06:13:12 Joshua So we also then started building these automated workflows where we could, because a translator can only do 2000 words a day. So we needed an automated workflow where we could split, you know, an 8000 word document into multiple, you know, sometimes 20, 30 people.

And we could have a team of specially trained editors working on it and return these within 40 minutes. 00:06:13:14 - 00:06:22:03 Joshua So this is human in the loop AI and automated workflows. Hopefully I'm getting your audience excited here. 00:06:22:06 - 00:06:28:22 Michael Yeah, it's interesting and a good background I think for the rest of the conversation.

So that's good. 00:06:28:24 - 00:06:54:00 Joshua Yeah. So that's how we got going. And then of course the Great Recession happened in the party ended, and I actually lived on Pine Street at the time, which is behind Wall Street, if, you know, an York City area.

And I could literally see people that I knew that I sold to carrying out boxes and I would bump into them on the street and, and they would be going. 00:06:54:01 - 00:07:21:01 Joshua And then what I learned, everyone else learned, is that George Bush was leading the government, and he had this idea to pump hundreds of billions, if not trillions, into the economy through public contracting, sorry, public sector contracting or government contracting. So it was a simple decision for a very young, simple mind to then pivot, to go to the government where the money is.

00:07:21:03 - 00:07:54:15 Joshua And fairly quickly, we found that those deals were lucrative, they were large, they were difficult and challenging, but they surprisingly were really happy to adopt the technology because they wanted to drive down price, at least at the unit rate, and they wanted to get things done quicker. And, you know, it was life or death. A lot of this stuff, it was no longer about trading, and they would value this tech in a way that the public, the private sector, funnily enough, wouldn't.

00:07:54:18 - 00:08:25:20 Joshua And it wasn't true with every section we used to sell to prisons. And I would go round mainly in the UK and they would use typewriters because they needed to air gap. But you know, so it wasn't true for all of public sector. So that's how we got in the public sector business.

And in 2013, I remained on the board of the Big Word, but I left to go and start a new company. 00:08:25:20 - 00:09:10:08 Joshua It was largely funded by the big Word, and it was a defense contractor, because what I was doing was very much providing a tech enabled service. But the service side that was human was still short term contracting. So I went into the contingent workforce business and applied all about what I know to a business that had no tech, and I thought I could disrupt it, which we did, you know, and government what they would do, especially with like top secret cleared and and human intelligence, they would have a small pool of people that would charge and say courts of $1 million for to be to operate abroad.

00:09:10:10 - 00:09:39:07 Joshua And I could find these people at 15 bucks an hour delivering Domino's Pizza because I could deploy technology to to look through social media and identify, oh, that's an Afghan. They've lived there over five years. I can't see any court records on them, a negative, maybe they got married or something. Divorce.

They're going to be able to get clearance and we can then deploy them. 00:09:39:07 - 00:10:12:03 Joshua So we would go and I used to remember Domino's Pizza guy $15 an hour. Would you like to $150,000. Go home and and and help you country.

And most would say no not for any money. Am I going back? But some would say yes. Yeah.

And you know, and we built this really great business in 2017, we merged the two companies back together and, and I took over as CEO of the whole group. 00:10:12:06 - 00:10:49:15 Joshua But we built a company from scratch to $20 million, generating around $3 million in cash every single year. And then in 2019, I had this vision that we were going to essentially be a multilingual communications business. We were going to be able to deliver video on demand video interpreters, humans, AI interpreting.

It was really easy then, because we had neural machine learning and we didn't have large language, but we as a business had a lot of corporate to train it in. 00:10:49:18 - 00:11:20:00 Joshua And so we could have AI, we could have human on demand remotely via video and over the phone. And we could also do it like Uber in person. And then you had all the written stuff.

So the subtitling, dubbing, editing as well with the new media services, and we were going to put it all in an application. We spent about $5 million building the user interfaces around the AI and the automation, and we had previously spent about $10 million. 00:11:20:00 - 00:11:45:09 Joshua So we were building on top of a system there. And just as it really started to happen, Covid hit and it hit really hard our industry, because 47% of all the work we did was like Uber style, you know, like you tap a button and there's a lot of automation happens, but eventually you get a human turn up and you can have a human in the hospitals.

00:11:45:09 - 00:12:10:06 Joshua Healthcare was a huge part of what we did in most of the world. Healthcare is public sector in government. That was pretty much our only private sector business in the US today. So we lost 47% of our revenue and I $200,000 a day of fixed costs.

So it was a scary time and I realized I had no rich uncle to call, sadly. 00:12:10:08 - 00:12:46:03 Joshua So yeah. So we we stabilize the business, we move more to digital. You know, necessity was a mother of all invention.

We created a really good video tools. And it's a long story, slightly less long. In 2021, we sold a majority stake to Susquehanna, which is a private equity fund run by by essentially Susquehanna International Group, which is the largest options trading bank in the world, I believe. 00:12:46:06 - 00:13:06:07 Joshua So, you know, we we kind of been through it.

And then you had LMS like a year later just as everything stabilized, came out and like, oh, I can do everything for free. And I think people have now gone on the same journey as I have since the 90s, realizing AI is spectacular. It can do all some things. 00:13:06:07 - 00:13:37:13 Joshua If you don't adopt it quickly and well, you're not going to be as cheap as your competitors and you're not going to be as quick as you competition, you're going to be able to automate quality.

However, it's not free. It's actually maybe not that much cheaper than humans, and often it's more expensive than humans. And, you know, we've been on that journey now with everyone else, except I think you probably have had a few more wins than I'm hearing out in the market space are operations. 00:13:37:13 - 00:14:04:16 Joshua Costs went from an average of $0.

18 in the dollar, down to around four and a half cent in the dollar. And that's a mixture of of of AI and just non AI workflows. So I call them self-driving workflows. But these are workflows that can essentially build themselves through optionality and and optimize for what the individual clients looking to achieve.

00:14:04:22 - 00:14:44:12 Joshua And then we also have been deploying AI interpretation which interpretations have spoken. Word translation is a written word. Most people just call it translation. But ultimately we always have the written word for a long time.

In 2013, neural machine translation came of age. It was actually developed originally in the 60s by I can't remember which university, but it was an open source program called Moses, and it was then became a hybrid AI program rather than a strict rules base. 00:14:44:14 - 00:15:09:19 Joshua And in 2013, Google put a lot of money into neural networks, and one of their big programs was machine translation. We worked with them on it.

They were a big client of mine for many, many years. They spent millions and millions of dollars on tuning it. So we've been through this and we we, you know, and each time you go through it, what it does is it explodes the market size. 00:15:09:19 - 00:15:25:00 Joshua So you do a much smaller percentage of a much larger market.

And in the math, in the language industry is it continues to grow by about 6% a year. 00:15:25:02 - 00:15:43:13 Michael Well, you know, in the in the prep call, we were talking a bit about the math around this. And I know you you've deployed a lot of agents in your call centers and and a lot of positive things out of that. Right?

I mean, they don't take days off. They don't have bad mornings. They, you know, they show up all the time. 00:15:43:14 - 00:16:06:07 Michael Right?

But the one thing you said that I thought was really interesting and surprising is that the those agents are actually more expensive than your human agents that they're supposed to augment. And I know you're running a call center in India. I'm just curious. Talk a little bit about the math and and why that's different than what most people think.

00:16:06:09 - 00:16:37:16 Joshua Sure. So I'm probably going to get myself in trouble sharing the secrets, but I'll do it anyway. If your listeners, we pay roughly about $5,000 a seat in the call center at salary level. What people don't realize about low cost locations is the Dell laptops they all have.

They cost the same, you know, as America. Yeah. And I know someone's going to say that 5% cheaper and they might be, but they're pretty much the same. 00:16:37:21 - 00:17:05:27 Joshua Yeah.

The building believe it or not. But the square foot cost for our building in Pune, India is actually more expensive than our call center in the UK. So the square foot they're sitting in and if you've run a call center, you'll know that at least 50% of your cost isn't the salary. So I've got 50% of the cost is the same.

00:17:06:00 - 00:17:35:28 Joshua The salary is much lower. You know that we pay a really high salary for India and we get really top quality people. I mean, our average call center person has two degrees, so they're smart, they speak English and they're very, very capable personnel. But ultimately my all in costs in India is somewhere around 10 to $12,000 per person.

00:17:36:01 - 00:18:07:20 Joshua Now, the AI, the license fee is around three and a half to $4,000. So on paper much cheaper. But here's the here's the reality. The AI can't do everything, and it's only as good as what you integrate into.

And that is a hidden cost. So we're still spending half $1 million plus on just maintaining the integrations into our databases. 00:18:07:20 - 00:18:32:28 Joshua Because if you want to call up and say cancel a face to face booking, it's got to be integrated in the back end of my database. And every time my database upgrades, I have to upgrade.

You know, it's on a three month upgrade cycle, one monthly patch cycle. So I have teams of people working on this and that then makes that AI significantly more expensive because I've got to keep redeveloping. 00:18:33:01 - 00:19:05:06 Joshua On top of that, I need a much larger infrastructure to run the AI. And, you know, from an IT perspective.

I've also got to tune the AI. So I have people doing that, and then I have to build orchestration tools because it's not one AI license, it's multiple, multiple versions, even sometimes you've just the same one. So again, once you actually look at it, the cost is very, very similar to a human. 00:19:05:07 - 00:19:31:08 Joshua However, the AI can only do about 70% of what the human can do about 70% of the time, which means there are about half as adequate net as a human.

Now, if I don't have them, then you're going to sit on my on the phone to my call center dispute payment for a long time. And, you know, so the service quality is much better. 00:19:31:10 - 00:19:54:15 Joshua And people tell me, Josh, but over time the price is going to come down. And I say, based on what?

Like I am a fax based guy, you know, based on what? Like what are you seeing? You know, and they'll say, well, look at televisions. And I say, but that's hardware that could easily be mass produced.

It didn't require a lot of electricity. 00:19:54:18 - 00:20:26:04 Joshua And here's the facts I see. I was just reading about it this weekend. The anthropic is signed last year, $1.

4 trillion worth of deals with invidious and all these different companies. Its revenue is only like $22 billion. Yeah. So think about how much money non CapEx it's losing about $12 billion a year which is like 50% of its revenue is loss.

00:20:26:07 - 00:20:48:01 Joshua Its growth rate is is spectacular which is why it's still worth a lot. But it's negatively scaling. And the math is always been by a lot of investors. Well it's in losing 12 billion.

If I can get it to say 50 billion it breaks even. But actually when you factor in the costs of these build outs, it's not true at all. 00:20:48:02 - 00:21:10:20 Joshua I mean, it's going it negatively scales. No one wants to admit it because they don't want to miss the train.

They don't want to be the idiot. They never invested when they had an opportunity. And there's a FOMO going on at the highest levels of tech society in Silicon Valley and around the world. So you've got that as an issue.

00:21:10:20 - 00:21:38:07 Joshua And then the big thing and I was just speaking to my head of IT and infrastructure, and he was actually telling me you were probably going to have to rebuild these data centers about every four years, because the amount of heat that's going through these and electricity and these chips don't last forever, and these servers don't last forever, and they're lasting way less than the experts predicted. 00:21:38:09 - 00:22:03:01 Joshua And they're under immense strain.

So if you think that that's like $1.4 trillion over the next few years. Current member. Exactly.

I think it was four years. What happens in another four years? You've got to rebuild that. So I'm buying AI at a massive discount right now.

And when I build my own, you know, I take I don't build AI, I build around AI and I take engines and I tune them. 00:22:03:01 - 00:22:21:22 Joshua But when I do that and I deploy it to my client base, my clients had charged about the same as a human, and that's the only way I can make money on it, because I can't burn billions of dollars. You know, I'm in the hundreds of millions, not the billions of dollars. So I can't afford to do that on their behalf.

00:22:21:22 - 00:22:43:24 Joshua And I don't have investors that would do that either. So you know why then deploy it. Why then have AI? Well, there's a couple of reasons.

It's instant it, as you say, it doesn't come in with a hangover or foul mood, or it doesn't commit a crime without you knowing about it when they're supposed to be, you know, an anti-terror cleared linguist. 00:22:44:00 - 00:23:09:27 Joshua You know, it's stable. It's just there. And it has huge upsides.

But I think we've got to be realistic about it. The integrations that you're going to need is what makes it magical. But they're very expensive to maintain. They need security patching.

You've got to wrap around the ISOs at your clients require all my clients require about 20 different ISOs. 00:23:10:00 - 00:23:46:28 Joshua You know, I'm a government contractor, so I have to have Cyber Essentials. Plus in the UK I have to be Hippo compliant in the US after the GDPR compliant in Europe, GDPR UK compliant in in the UK. And you know I have to have server farms you know to to manage this and orchestrate all this data.

I have to have it it teams doing quarterly business reviews with every single supplier that's supplying, doing the questionnaires, making sure they're compliant. 00:23:46:28 - 00:24:03:19 Joshua I have my clients do the same to me. I mean, we do over a hundred audits from clients a year, so I've got to pay for all of that. Even when they're using.

I can't be free. And I think people, because they're so used to paying 20 bucks a month for ChatGPT or nothing. 00:24:03:24 - 00:24:05:12 Michael Or nothing. Yeah.

00:24:05:14 - 00:24:26:24 Joshua You know, they think this is free. So, yeah, the good news is if you're a human, you can still compete today with AI. And until they figure out and I, I think that they've deployed a lot of money to try and do so how they can run these data centers more efficiently and cheaper and how chips can get cheaper. 00:24:26:24 - 00:24:53:12 Joshua And they will eventually figure it out because there are small incremental improvements.

But I think that the AI revolution is going to be more like a decade or two. I don't think it'll be like the internet revolution, which I believe the internet revolution took three decades, really going back from the 80s to say 2010, before we were really utilizing it to its best of its ability. 00:24:53:13 - 00:25:08:06 Joshua I think this might be around half of that, but there's a lot of time to figure this out. And I I've got a feeling that by the time the revolution has ended, it's just be a is what we would call as business as usual.

00:25:08:10 - 00:25:36:18 Michael Yeah. I mean, you know, we we start to hear there's been a lot of excitement around obviously. And we're starting to hear companies talk about ROI bit, but the numbers aren't really playing out yet. And I mean, obviously the the infrastructure cost, everything is still very expensive and it hasn't come down yet.

I mean, I guess it will eventually, but but you know, I'm really curious about the, the the kind of the focus. 00:25:36:19 - 00:25:59:28 Michael Right. So we focus on the, the adopting it. We have to have it.

And it seems like we're not really focusing that much on what's it doing, what's the out, you know, what are we trying to do with it. Right. What's the outcome. I mean, how did that happen?

And then, you know, how do you how do you think companies can get the balance right between the investment you have to make and then what you're going to get out of it in the short end? 00:25:59:28 - 00:26:00:27 Michael The long term? 00:26:01:00 - 00:26:40:21 Joshua Yeah. So I'll deal with us.

How did it happen? I don't know who said it, but we say a lot in America. Follow the money. So you have these huge companies.

Anthropic, OpenAI. They didn't exist that long ago. And they're raising billions in videos like a computer game, like niche chip manufacturer. Right.

You know, so they're raising a lot of money by convincing everyone that this is a revolution. 00:26:40:21 - 00:27:04:21 Joshua You get people like Elon Musk, no one will have a job. We need universal salaries for everyone that the government will have to pay. And what no one's questioning it because these are successful people that have done well, that seem to know what they're talking about.

What you may not realize is you've been neurologically programed. Your programing is the science behind brainwashing, but is real. 00:27:04:22 - 00:27:28:15 Joshua It happens. That's how you can have propaganda and it still works today.

But everyone is being programed to believe that AI is coming is going to take over. We don't know exactly how it's going to take over. No one's asking that question. That's why these podcasts are Greeks.

The mainstream media aren't really asking it. No one's asking. Is it going to be profitable or how does it scale it? 00:27:28:18 - 00:27:54:28 Joshua Those those aren't being asked because they're inconvenient, because all the billionaires, the control the media and own it, they've got their money stuck in these businesses and they need you to keep retail buying to pump up the price.

So I know it sounds like a conspiracy theory and maybe it is, but you know, but I'm seeing this. So as a CEO, I can tell you that we have had and this is again, this is anecdotal. 00:27:55:01 - 00:28:19:15 Joshua I'm sure there's research that I could have done before I came on with you, but we have had one of the greatest firings of CEOs over the last 12 months that I remember. And most CEOs that I've met have been fired.

And why have they been focused? Their boards are insisting on deploying AI, and the CEOs are asking the questions that you're asking. 00:28:19:15 - 00:28:40:28 Joshua Well, I need to figure out what I'm trying to solve this. I don't want to create a solution looking for a problem.

I need to ultimately invest smartly because most CEOs have been through this before, many, many times. The average board member not so much because the last time this happened was the.com boom and bust. And that was in the 90s.

00:28:41:00 - 00:29:03:26 Joshua And most board members now were very young. Then they went on boards so that they'd been fired. And this is a fact that I read that 80% of CEOs that are hired or over 80%, I believe it is actually first time CEOs. And, you know, I believe that's because people are saying, you know, I'll put my hand, I'll give it a go.

00:29:03:28 - 00:29:23:24 Joshua And boards are desperate because they believe that AI is coming to replace humans, and that my investment isn't going to be worth anything unless I can deploy AI. And if I can, maybe I can capture market share early and so on. So there's a lot of hysteria. There's a lot of panic.

It's not really based on data, it's not really based on facts. 00:29:23:24 - 00:29:43:00 Joshua And then you get the people like you, Michael, who are out there trying to get to the bottom of some of these challenges. And I think most people come on your podcast saying the same thing. AI is great.

It's going to be revolutionary, but it's way harder, way more expensive than anyone thinks. It's going to take way longer. 00:29:43:02 - 00:30:03:25 Joshua No one's listening to us yet, you know, because we're not on the mainstream media. If this was Fox News or CNN, you know, we wouldn't get on it because no wants to hear the the reality.

They want to live in la la land for a little while now. That's dangerous, because what will ultimately happen is there'll be the counter to this. 00:30:03:27 - 00:30:29:02 Joshua You know, the the people who bet against this, they'll become sexy because they short the market and they'll make movies on them. About the big short of the 2028, you know, Great Recession or whatever is about to come.

And they'll ignore the fact that it was nothing to do with AI. It was probably to do with the fact we've had a bull market for the last 15 years, and it's time to correct. 00:30:29:04 - 00:30:53:16 Joshua But that's how this is happening. You're getting very clever people who understand that they can make more money through talking it up than actually delivering.

And right now I don't see I mean, Microsoft is a big partner of the big words, and I love Microsoft and I think they're great. But when I read that data, it almost looks like they're losing $4 to every $1 they're spending on AI. 00:30:53:19 - 00:31:07:09 Joshua And I'm thinking, well, I like partnering with them because I'm getting all this free tech essentially, you know, 75% discount on their cost to make it. But I sure as hell wouldn't want to be an investor.

00:31:07:12 - 00:31:29:21 Michael Yeah. You know, it's interesting over the last few months, of course, the the the explosion around. Oh come now for the SaaS industry. And we've seen this as a collapse.

And I've been writing a lot about, you know, let's wake up. The reality is that these enterprise applications are not going away. You can't do business without the data and the business logic and all these other things. 00:31:29:21 - 00:31:54:20 Michael Right?

The AI doesn't replace that. It does other things that are valuable. Right. But you know, something you mentioned, it really jumps out to me from a CEO perspective.

We are going to see a lot of failures over the next year or so, right? I mean, we're we're applying AI in situations where we don't have clear outcomes and, and we don't really even have understanding of what the process is. 00:31:54:21 - 00:32:19:27 Michael So when things fail, it's going to be seen as a big setback and the boards are going to lose even more confidence. Management team is going to lose confidence, right?

I mean, what how should companies be thinking about, you know, how do you stage your deployment to protect against this confidence collapse? That really could be catastrophic for for some companies? 00:32:19:28 - 00:32:35:04 Joshua Yeah. I mean, I can tell you I sit in these boardrooms.

I'm also an investor in business. It's already happening. You're already getting very, very frustrated investors saying we've just spent millions of dollars on building out this AI. 00:32:35:07 - 00:33:00:09 Joshua Where's the return on investment?

And I guarantee you that's happening in, you know, chat board or OpenAI, should I say. And there's going to be some big winners and there's going to be a lot of losers. I mean, that was the same as the.com boom and bust.

You know, people forget that we we have eBay today thanks to the.com boom we have Google today. 00:33:00:12 - 00:33:37:19 Joshua Even Gemini exists only because of the.com boom and the money that flowed into that.

And that's what happens. There are companies that had real opportunity because they existed before the AI, before the automation. They existed to solve real problems. And they're now deploying this new capability in new technology to solve those problems cheaper and quicker.

And typically those that can do it and they can do it profitably are going to be huge winners. 00:33:37:21 - 00:34:03:16 Joshua And then you're going to have all these companies that basically integrated with ChatGPT AI and didn't really add any value or solve the real problem. You know, I'll give you an example. It's probably a bad example, because I think Facebook will ultimately be a winner of this, but they started putting AI into WhatsApp.

I don't know if you use WAP WhatsApp and I go, but it's so annoying. 00:34:03:16 - 00:34:32:16 Joshua I just want to search my messages. I don't need an assistant. No, no, I have Kyle, who's my real human assistant, and I have Gemini, who's my AI assistant, and I have open AI.

That's my other sister, and I have anthropic trying to tell me how to do my job better every five seconds. I'm like. And sometimes I just let Copilot and anthropic argue it out. 00:34:32:19 - 00:34:57:05 Joshua I don't need a WhatsApp assistant in my, you know, personal messages telling me.

And then I drive a Tesla and all of a sudden grok keeps popping up every five seconds, you know, and, you know, at least grok solve the problem. I've got kids and sometimes will be in a long car ride and they'll do, you know, like rocks. 00:34:57:06 - 00:35:16:14 Joshua Give me some questions and they'll try and answer them and then they'll argue with grok as I've got girls. So they'll argue of who's right and who's wrong on the lunar cycle or whatever.

But, you know, we we you know, a lot of this AI doesn't really solve a problem. It's a solution looking for a problem. And I think that's what's going to happen. 00:35:16:15 - 00:35:47:08 Joshua Michael, I think you're going to find that people who are well funded, who haven't spent a lot of money for the sake of spending it, people who are disciplined investors who may not be like the big word, we're not seen as a sexy AI company.

Everyone tells me you're not real tech company. And you say, okay, okay, but I've been around 44 years and gone through three of these revolutions, and I'll be around in another 44 years going through way more. 00:35:47:10 - 00:36:14:09 Joshua You know, I always say the big word and I say this in most of my conferences internals, we need to be cockroaches. Now, people think negatively of cockroaches, but I don't they're not beautiful.

And the big word isn't always beautiful. But we can survive a nuclear blast. We just. How do?

Well, in the most uncertain environments. But why do cockroaches do well? 00:36:14:10 - 00:36:34:21 Joshua Because the built for that and the big word is built for that. And the other companies are going to be built for that as well.

And they're going to do well. Most of them though, you know, I was asked to invest in a company that makes AI for banks and I, you know, and they were bragging to me that they built this algorithm on a weekend. 00:36:34:24 - 00:37:00:02 Joshua That's how good they are. And I was thinking, well, it's naive to think that there isn't someone in the banks that can do the same thing.

I mean, these are big banks, so why would I invest in something that just takes a weekend to build? So, you know, but yet they'll raise tens of millions of dollars and, and maybe they'll deploy that cash and other a lot of value and I'll regret passing on it, but a lot of them won't. 00:37:00:04 - 00:37:00:22 Joshua Yeah. 00:37:01:00 - 00:37:26:22 Michael I you know, this year I've been thinking a lot about, you know, we're finally seeing companies move from proof of concept and pilot into production.

Right. And and it strikes me that one of the things and frankly, we've done this with governance for years. But one of the ideas here is that we're trying to, you know, we move into production and then we try to slap governance on the on the back end of it. 00:37:26:22 - 00:37:51:19 Michael Right?

Oh, no, we made a mistake. How do we do this? And and and because we're, we've we're moving at machine speed now things can happen much differently than they used to. And you know, I think I've been writing about governance by design.

I think this is really important. We have to think about how we shift that model and put the governance in up front versus after the fact, because after the fact, frankly, when you're dealing with AI, it's just too late, right? 00:37:51:20 - 00:38:11:06 Michael And I know you have a lot of experience in highly regulated environments from you mentioned HIPAA and government regulations, that sort of thing. I mean, what do you think governance should look like when the stakes are that high?

And, you know, speed is is it could ultimately be your enemy as well as your friend? 00:38:11:08 - 00:38:34:20 Joshua Yeah, I think it's a good question. A lot of people don't know when they when they hear people talking about governance of what it really means. I think they know it, technically speaking, that governance is governing and governing how something should be done.

But, you know, so I'll try and break it down for your audience, you know, into simple buckets. 00:38:34:20 - 00:39:03:08 Joshua So the first thing is, is that your nonnative business has governance, even if it's unofficial governance. So, you know, what are your operating procedures to stop a catastrophe or stop data falling into the wrong hands? So you'll have data security governance, you'll have cybersecurity governance, but you'll also have very human governance.

So, you know, HR rules, you know, and we have HR laws. 00:39:03:08 - 00:39:33:09 Joshua It's all governance. And I think the way we make this new world work is, and what you're pointing out is when you get it wrong in the human world, it's pretty slow. And it doesn't go wrong 30,000 times in the same three milliseconds.

But when it does that in the AI world where, you know, for example, in my call center, one AI agent can call 30 people concurrently and have many AI agents. 00:39:33:09 - 00:39:57:15 Joshua And if if that call goes horribly wrong, it's times 31st is the human world where it's times one. So but what you have to do is you really have to treat your AI as human, and you have to realize that the margin of average has got way smaller. And I think that's that's what we do at the big Word.

00:39:57:18 - 00:40:21:22 Joshua Our testing facility now is vast, and the amount of people we have in that testing facility. And the cool thing is you can use AI to do testing as well, but you don't want, you know, the AI policing the AI without humans. And I sort of joke on LinkedIn, it was a cartoon and it said, why aren't you doing anything? 00:40:21:22 - 00:40:46:16 Joshua And the guy had a QA title and he says, because the AI is doing it for me.

And, you know, and I think you have to be careful on that because, again, we underestimate human intelligence. My wife was telling me the other day how she knows someone that's having a relationship with AI. So I said it's a superficial relationship. 00:40:46:18 - 00:41:18:02 Joshua And she was a little irritated with me that I said that and she said, no listens.

It's really nice. So I said, I know that it's a potential superficial relationship, and for a little while that works. And you can con yourself into believing that this is, you know, human like intelligence or human intelligence, but in artificial form. But in reality, that relationship was the best it could ever be on day one.

00:41:18:02 - 00:41:57:16 Joshua It could never get better. And what we as humans are underestimating is the is our capability. That actually cannot be. In my opinion.

I might be going to spiritual here, Michael for your podcast, but cannot be fully replicated. It can be largely and you can trick people for a little while. And I think in terms of governance, just going back to your question, we need to have as, as organizations, really strict rules on how we govern data now, how we the processes to deploy AI. 00:41:57:19 - 00:42:35:01 Joshua There's a there's a race out there.

Everyone's trying to be first and beat their competition. But this is where the tortoise will win the race and the hairs will be the.com busters. You know it'll be the AI busters soon.

EBay, Google, PayPal. They were real, you know, solutions to really difficult problems. You know, machine translation came along in our industry, and all of a sudden you could translate your website and be international, even though you had no physical location anywhere other than your own bedroom. 00:42:35:02 - 00:43:04:20 Joshua But what people didn't realize is we don't have logistics, so I can sell my antiques to China, you know, from my little bedroom in Minnesota.

But if I can't get it delivered for any less than $20,000 because I've got to rent a plane or a ship to get there right then I can't, you know, and I think there's a version of that going on with AI that we haven't built all the infrastructure around it yet to make it truly safe. 00:43:04:20 - 00:43:30:06 Joshua So the people who are deploying it now quickly, they're taking a big risk. Some of those risks will pay off, of course, but in our business we're very methodical and it's slow and it's hard and it's expensive as well.

But I believe that in ten, 15 years time, the big word will be doing exactly what it's doing today with, of course, we won't even call it AI. 00:43:30:08 - 00:43:38:10 Joshua It'll be like when remember when you call it the World Wide Web, you know, we don't call it a wide web. It's just the net. 00:43:38:12 - 00:44:11:01 Michael Yeah, it's just the net.

Yeah, it's. It's like that old Intel commercial. You know, the little sticker on your computer says there's Intel inside. You know it's there.

But do I really think about it? No. Yeah. That's interesting.

Well, I mean, you've been through a bunch of waves of change with AI through through the years. And, you know, I know that we're in another one of these hype cycles, but but from your perspective, what's this longer view taught you about how industries actually absorb change like that? 00:44:11:01 - 00:44:22:03 Michael And then what do you think the next few years look like for enterprise AI and and for business as they try to move through this, this cycle? 00:44:22:06 - 00:44:55:03 Joshua Yeah, it's a good question.

I would encourage your viewers to really watch the high risk, regulated industries, because they're going to be the ones that smarter investors. So watch the drug manufacturers. They have so much upside in AI, but they're not bringing out cancer drugs every few weeks like people thought two years ago. Why?

Because they are deploying AI and they're going to get there, but they're going to get there methodically. 00:44:55:06 - 00:45:18:14 Joshua So if you want to symbol answer to the question is, watch the people that do this at scale, watch the big words, watch how we slowly bring it out and and the use cases that we use for AI versus humans. So for example, in the court system, we're still going to use humans to interpret witness statements because it can't be 97% accurate. 00:45:18:15 - 00:45:48:26 Joshua It's going to be 100%.

Imagine being accused of murder. And the witness is only going to be maybe 97% accurate, but also could be 85% accurate. You know, it would be thrown out of court. So we have to, but we're going to be able to use it on the triage side.

So, you know, before that witness shows up and they're getting instructions, attend the court at 10:00 and you're going to check in here and you're going to go upstairs here and that it's great. 00:45:48:26 - 00:46:08:01 Joshua And then they're going to be on time, and the courts are going to run on time. And the witness isn't going to be terrified about being on the stand in a murder trial. Same in hospitals.

You're going to have a human with your doctor, but you're only going to have a machine explain how to get to the doctor, because it's on the third floor and it's the third room on the right. 00:46:08:01 - 00:46:37:26 Joshua And, you know, for the blood test, maybe you're going to get an AI interpreter because we don't really there's nothing that crazy about a blood test you don't really need to know about much. But when it comes to aftercare and you've had you call an Oscar B and they want to say, you know, if you see blood in your stool called immediately, well, that's what life or death, internal bleeding situation and you know, so watch the companies that do this deployed it before and get it.

00:46:37:27 - 00:47:06:08 Joshua You'll get a really good idea of what's going to happen. And I think it's going to take a lot of time, but eventually you're going to see drugs come out much quicker as we use AI models to predict what's going to happen, which then we can apply to the mice and the human trials. And there's still going to be human trials at the end of the day, you know, you know, and I, you know, in the the cycle of technology goes on. 00:47:06:08 - 00:47:15:00 Joshua And this is just another to your point, there's a lot of hype.

But under that hype, AI is real and it will revolutionize the world. 00:47:15:02 - 00:47:37:20 Michael Yeah. I mean, I certainly am using a lot in my business. And I see, you know, with my clients, I talked to a lot of people that are really seeing value out of it.

But but I also talked to a lot of people that are looking for value in the wrong places. Right? They're not they're not actually thinking about what the business outcomes are and how you actually get to that point and get what you want out of it. 00:47:37:21 - 00:47:38:16 Michael Yeah, yeah.

00:47:38:18 - 00:47:59:21 Joshua And the unintended consequences. You could tune an AI version of yourself. You got a lot of podcasts, you could put a podcast out a day. EIU people are going to get blind to that noise.

It's going to be repetitive because it's not learning more. It's just learning of itself, you know? So there's there's going to be a lot of unintended consequences. 00:47:59:21 - 00:48:10:14 Joshua And I think one of them will be data saturation product and service saturation across different industries.

00:48:10:16 - 00:48:20:03 Joshua That don't really offer a lot of value. But because it's cheap, people will buy it at first until they realize there's just so much of this stuff. I'm not interested anymore. Yeah.

00:48:20:04 - 00:48:40:07 Michael Yeah, that makes sense. Well, at the end of every episode, I like to ask the guest to, to to offer something to the, to the audience of value. So some recommend someone who, you know, a thought leader, an author, podcaster, somebody that you think the audience would, you know, enjoy and get a lot of value from. 00:48:40:09 - 00:49:04:14 Joshua Yeah.

So I'm a bit biased here, Michael. I've done admit, but my father has a podcast and he's an entrepreneur who sold two businesses. He never did it for the money. He made a lot of money over his time.

But he's an insatiable investor. He's not a techie, but he really understands the power of technology in a way that most people in the weeds can. 00:49:04:15 - 00:49:27:02 Joshua He wrote a book called I think it's called The Making of Mr. Irresistible.

He came from government housing with no money left. School at 15, had a pretty difficult life and built, you know, businesses forever and sold them for over $100 million each. In the 90s he had a business that was dead. He sold it, did 24 million to Mike Mulligan, the junk bond billionaire.

00:49:27:06 - 00:49:39:28 Joshua And it's just a really good real story without hype. So I encourage you to watch gold with Gold. And it's he does that with Cornell University. 00:49:40:01 - 00:49:50:25 Michael That's super.

Yeah, that's a great recommendation. I'm going to have to check that one out myself too. Joshua, thanks so much for joining today. Really, really useful and interesting conversation.

00:49:50:27 - 00:49:59:01 Joshua Great. Thanks a lot. Michael, and great to be on your podcast. Thanks.

00:49:59:03 - 00:50:23:02 Michael And that's the show for this week. Thank you all for joining us. Remember to like, share and subscribe to the show. If you enjoy the show, please leave us a review to help others find us.

For more research on AI and other software, check out Arionresearch.com If you're an expert in AI, generative AI, or business automation, either as a provider or an inducer, email your information to disambiguation at arionresearch.com 00:50:23:04 - 00:50:31:21 Michael Don't forget to join us next week! Disambiguation is an area in research production.

I'm Michael Fauscette and this is the disambiguation podcast.