Oct 21, 2026 · 37 min · 12 segments
In this episode of Tales from the Pros, Eric Lawrence sits down with Matt Soltau, Global Director of Strategy and Operations at IntelliPaaS, to discuss why so many AI projects struggle to move beyond…
Matt SoltauGuestEric LawrenceHostMichael GeorgiouHostAnd it's interesting because you're starting to see a lot of companies start to realize that the cost of what they're spending on AI is bigger than the impact.
And I'm interested to get your take with the vast majority of AI pilots that deliver zero business impact.

Yeah, look, I mean, how can I put this? The graveyard, AI graveyard is ultimately full of projects that actually didn't fail at the model, but failed at the pipe that feed it because pilots often collapse before they make it outside of their sandbox.

But then once you release that into the wild world, the real data that is then being relied on is often dirty, it's duplicated, it's untraced, and that often becomes the killer for it.

It's getting quite late, but they've had a particular well, ultimately, they try to deploy an AI chatbot to advise their team real time on the phone with incoming customer inquiries.

Following scenario, you're calling your insurance company, that insurance company has real people on the phone, but they may not be fully trained in everything, right? Because it's quite a few different insurance topics that they have to cover.

So now transcript runs in the background and then the AI shows in real time what the agent is supposed to be telling the customer.

Sounds pretty good in principle, worked really well in the sandbox environment, ultimately failed in production.

Why? Not because the model itself was bad or the idea was bad, et cetera, like great idea.

And in the end, we see stories like this all the time where you've got on the one side a fantastic idea that gets tested, works really well.

So initially, you've got your new puppy dog that is trained to maybe find its way around the house.
And it's interesting because you're starting to see a lot of companies start to realize that the cost of what they're spending on AI is bigger than the impact.
And I'm interested to get your take with the vast majority of AI pilots that deliver zero business impact.

Yeah, look, I mean, how can I put this? The graveyard, AI graveyard is ultimately full of projects that actually didn't fail at the model, but failed at the pipe that feed it because pilots often collapse before they make it outside of their sandbox.

But then once you release that into the wild world, the real data that is then being relied on is often dirty, it's duplicated, it's untraced, and that often becomes the killer for it.

It's getting quite late, but they've had a particular well, ultimately, they try to deploy an AI chatbot to advise their team real time on the phone with incoming customer inquiries.

Following scenario, you're calling your insurance company, that insurance company has real people on the phone, but they may not be fully trained in everything, right? Because it's quite a few different insurance topics that they have to cover.

So now transcript runs in the background and then the AI shows in real time what the agent is supposed to be telling the customer.

Sounds pretty good in principle, worked really well in the sandbox environment, ultimately failed in production.

Why? Not because the model itself was bad or the idea was bad, et cetera, like great idea.

And in the end, we see stories like this all the time where you've got on the one side a fantastic idea that gets tested, works really well.

So initially, you've got your new puppy dog that is trained to maybe find its way around the house.
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.