TECHtonic: Trends in Technology and Services
Aug 21, 2026 · 33 min · 10 segments
What happens when an AI agent fails, and how can enterprises prove they saw it coming? On this episode of TSIA’s TECHtonic, Thomas Lah takes on one of the most important questions facing organizations…
Sekhar SarukkaiGuest
Thomas LahHost
So playing it back, what you're saying is a lot of companies have probably pretty solid governance on the front end when they're deciding what tools to bring in to, to deploy, et cetera.
But once they've made that decision and the tool's up and running, it is the runtime governance that you're most concerned about.
And also, you know, to put this in perspective, I mean, historically, you choose a SaaS tool, CRM system, whatever, you deploy it, and all the front-end decisions around that, the governance, et cetera, the security, you know, super important.
And I think what's interesting is that you've said that agent failure can look chaotic, but it actually falls into very repeatable patterns, and that's a real claim, right? That agent failure can be put into a taxonomy.
No, let me sort of talk about that in the context of a, a completely different domain.
So if you go back a hundred years in the medical industry, I think in the nineteen-forties, the International Statistical Institute had a problem, which is doctors were sort of diagnosing various symptoms and converting symptoms to diseases, whatever.
But there was no common lingo or taxonomy by which they were able to communicate what they're doing.
So they came up with what's called the International Classification of Diseases.
...all the codes which represented how a human died, right? And it had a hundred and sixty-one categories.
It could be pneumonia, it could be some liver disease, whatever, right? And that standardization actually, uh, unlocked a lot of interesting side effects, right? The first thing is doctors were now able to reason about diseases and sort of reuse approaches to remediate, um, their patients.
Research was unlocked because now they were able to look at data across different practitioners.
So playing it back, what you're saying is a lot of companies have probably pretty solid governance on the front end when they're deciding what tools to bring in to, to deploy, et cetera.
But once they've made that decision and the tool's up and running, it is the runtime governance that you're most concerned about.
And also, you know, to put this in perspective, I mean, historically, you choose a SaaS tool, CRM system, whatever, you deploy it, and all the front-end decisions around that, the governance, et cetera, the security, you know, super important.
And I think what's interesting is that you've said that agent failure can look chaotic, but it actually falls into very repeatable patterns, and that's a real claim, right? That agent failure can be put into a taxonomy.
No, let me sort of talk about that in the context of a, a completely different domain.
So if you go back a hundred years in the medical industry, I think in the nineteen-forties, the International Statistical Institute had a problem, which is doctors were sort of diagnosing various symptoms and converting symptoms to diseases, whatever.
But there was no common lingo or taxonomy by which they were able to communicate what they're doing.
So they came up with what's called the International Classification of Diseases.
...all the codes which represented how a human died, right? And it had a hundred and sixty-one categories.
It could be pneumonia, it could be some liver disease, whatever, right? And that standardization actually, uh, unlocked a lot of interesting side effects, right? The first thing is doctors were now able to reason about diseases and sort of reuse approaches to remediate, um, their patients.
Research was unlocked because now they were able to look at data across different practitioners.
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.