Talking AI

The episode contains substantive discussion of agent vs. deterministic workflows, model evaluation methodology, token optimization, and organizational AI implementation strategy. However, there is notable filler including a mid-roll ad break, softball questions about daily recaps, and repetitive framings that could have been more concise. The best insights cluster around Automation Bench findings, hybrid agent-workflow architecture, and institutional vs. individual AI adoption - genuine operational concepts - but these are interspersed with generic advice.

“So you have to think about what are the things that these frontier labs can't do or they want to. And you really just try and understand what are the things that are really valuable to customers that those companies can't fulfill.”

“The best evals are typically evals that are actually kind of easy for humans to do, but really hard for these models to do because now you're testing something really interesting.”

Wade offers some genuinely fresh framings - the deterministic-vs-probabilistic workflow distinction is well-articulated and the Automation Bench benchmark result (18.1% success on real workflow tasks) is concrete and counterintuitive. His distinction between 'floor raisers' (broad organizational fluency via hackathons) and 'ceiling raisers' (institutional AI bottleneck-breaking) is useful. However, much of the narrative follows familiar patterns: the Code Red story is now well-known, and broader points about model selection, hybrid systems, and organizational transformation are circulating widely in enterprise AI discourse. Limited true contrarianism.

“The best model right now is performing these tasks only 18.1% of the time. So the best model right now is performing these tasks only 18.1% of the time.”

“You have this like, hybrid agent workflow setup and you want something like Zapier to sort of manage those processes end to end.”

Wade Foster is exceptionally well-calibrated for this topic: founder and 15-year CEO of a $5B SaaS infrastructure company directly affected by AI commoditization, early mover on organizational AI integration, and someone making real-time product decisions in the space. He speaks from operational experience, not theory. His credibility is reinforced by concrete examples (Zapier's employee count vs. AI agent count, Automation Bench data, actual product pivots). This is a practitioner at the right level and scale for B2B AI strategy discussion.

“Zapier has more AI agents than it has employees.”

“Wade Foster co founded Zapier in 2011 and he has led it ever since, turning a scrap UI Combinator startup into the $5 billion plumbing of the SaaS era.”

The episode contains strong specific data points: Automation Bench's 18.1% success rate, the Code Red hackathon moving employee AI adoption from 11% to 50% in one week, 600M automated tasks, and references to Coinbase's 50% token optimization. Concrete workflow examples (daily recap journal, email drafting) are included. However, many claims lack granular support: no details on the actual 'rethinking' that occurred post-Code Red, vague references to 'most sophisticated customers,' and limited financials on how Zapier itself is monetizing AI. The institutional AI section relies heavily on conceptual framing without case-level specificity.

“We saw our daily usage of AI from our employee base go from about 11% of folks using it to over 50% in one week.”

“And as you can see if you look at the leaderboard, the top of the leaderboard right now is GPT 5.6 SOL max and it's scoring 18.1%.”

The host asks reasonable setup questions and does pursue some follow-ups (e.g., on Automation Bench, on model efficiency), but overall the interview lacks edge and genuine productive friction. The host is respectful and knowledgeable but rarely presses back or challenges claims. Questions tend to invite Wade to elaborate on pre-formed talking points rather than exploring tensions or testing assumptions. The daily recap discussion, while interesting, goes largely unchallenged despite being somewhat generic productivity theater. The conversation meanders and doesn't drill into the hardest questions around organizational change resistance, financial impact of AI transformation, or Zapier's actual business model shifts.

“What was going through your head at that moment in time? I think it was a seminal moment for everybody, but I think few leaders took it to that extreme and really saw what was coming.”

“Yeah, and there's several rabbit holes I want to go down here.”