Impact of AI: Explored podcast, with James O’Regan and Gerjon Kunst · June 6, 2024 · ~60-minute conversation.
Brian’s first podcast appearance as an independent consultant, coming just before he joined Citrix. The episode is built around two pieces of news from that week — GPT-4o and Microsoft Recall — and turns into the first full public airing of what Brian starts calling “workplace AI.”
What Recall actually is, and what it enables
Microsoft Recall takes a screenshot of a Windows machine every few seconds, then uses AI to build a searchable index of everything that ever crossed the screen — “what was that website I was looking at last Tuesday” becomes an answerable question. The privacy implications are enormous on their own, but the corporate angle is sharper: run this on VDI desktops and you get an AI-powered compliance engine watching every employee’s screen continuously. PII gets blurred automatically. Data gets masked in insecure locations. None of this is hypothetical — the technology exists today.
The fiduciary-duty argument
If a business owner can use AI to record every employee’s screen, feed it to an analysis engine, and triple productivity — whether that means triple the revenue, a third the time to market, or the same output with a third of the headcount — a public company arguably has a fiduciary duty to shareholders to use it. Not using available technology that dramatically improves return on capital is itself a kind of dereliction. The trade Brian sketches: work devices, everything recorded, everything feeding a company-wide “superbrain,” in exchange for double the salary. Skepticism that employees would revolt gets met with a simple counter — there would be a line of resumes out the door for that deal, because people already accept far worse trades for far less money.
The productivity measurement problem
The old joke about the mainframe repairman: a machine costing $10 million a day in downtime gets fixed with one kick, and the bill reads “Kicking the mainframe: $10. Knowing where to kick the mainframe: $999,990.” AI-driven monitoring only sees the part of a person that touches a keyboard. It can’t see the hundred minutes of thinking that made the one-minute email possible, or the walk that made the idea click. Companies are going to be tempted to rank employees by screen output anyway, because that’s the only signal AI monitoring can actually capture — and it’s exactly the wrong signal for knowledge work.
What happens to the concept of work itself
If generating text output has zero marginal cost, the deeper question isn’t “how do we monitor this” — it’s “what am I doing all day.” Does everyone work half the time for the same money? Half the time for half the money? Do companies quietly stop backfilling roles as people leave rather than announcing mass layoffs, with the pain landing hardest on the college graduates who would have filled those entry-level seats? Nobody in the industry conversation is actually engaging with this. The public discourse stays at the level of “AI is going to change things, so make sure you do it right” — true, and completely unactionable.
The project this became
Even if AI capability froze at exactly its 2024 level, the products that already exist are enough to fundamentally change how organizations operate, how they relate to employees, and how they think about pay. That gap between what’s technically possible today and what anyone is seriously grappling with is what Brian describes organizing his own thinking around next — the same way he’d previously built out full books on VDI and DaaS.
Key formulations
“Knowing where to kick the mainframe: $999,990.”
“If I can triple your productivity, I’ll double your salary.”
“It’s all platitudes… that’s not actionable. What do I do with that?”
“Even if all AI evolution stops today, everything that exists as products today can really fundamentally change how organizations operate.”