Support has always been a launchpad for talent moving into product, sales, and marketing. Clay's George Dilthey stopped fighting it and built a formal program around it instead.

George Dilthey leads support at Clay, where his 25-person team handles 10,000 monthly tickets. Clay is a workflow automation platform that combines data sources and AI to help businesses scale manual processes. Before joining Clay in March 2024, Dilthey built support teams that became accidental talent pipelines—people who mastered the product and customer needs, then migrated to other functions.

The counterintuitive insight: Rather than fight natural attrition from support to other functions, Clay built "The Wheel"—a formal rotational program where new graduates spend 70% of their time in the support queue and 30% rotating through other departments, with the explicit goal of training expert generalists who'll eventually move elsewhere in the company.

Overhiring Support to Feed the Company

Most support teams carefully match headcount to ticket volume. Clay does the opposite. They deliberately overhire on support, treating it as a 6-12 month training ground that exports talent to marketing, sales, operations, and product teams.

The program started small—Dilthey and a recruiter built a Clay table hooked to a Typeform, skipping the formal ATS entirely. At their first NYU hiring fair, they printed 50 flyers and ran out in the first hour, maintaining a line at their booth for four hours straight. They were the only tech company at an event dominated by JP Morgan and the Department of Transportation.

This revealed something unexpected: universities desperately want tech company engagement. Career centers are highly incentivized to help students find jobs and will actively facilitate connections—but most tech companies aren't showing up to these fairs anymore. Clay found essentially zero competition for technically capable graduates who were hungry for any structured entry point into tech. The talent pipeline problem wasn't a supply issue; it was a distribution issue.

Clay's bet is that as AI handles specific tasks like coding or ticket resolution, the valuable skill becomes bridging across functions—exactly what support-to-rotation provides. Schools proved so eager to help that Clay now works directly with campus clubs and career centers, with one wheelie (a recent graduate herself) leading campus recruitment to build relationships that could produce future founders.

The Safety Net: High-Slope Hiring with Protective Structure

Clay doesn't just hire anyone willing to rotate. They specifically target what Dilthey calls "high-slope" talent—people with strong technical foundations who can become product and customer experts rapidly. This careful selection provides a safety mechanism: even while experimenting with rotations, the support team's core metrics improved. When the June cohort started, response times halved.

The rotation structure itself contains protective guardrails. The first three months are pure support—no rotations, just deep product and customer immersion. Only after that foundation do people move to 70/30 splits, ensuring support quality never depends on distracted part-timers. This staged approach means Clay can experiment with rotation placements without risking customer experience.

Current rotations span UX research, chief of staff projects (like building programs to help other companies hire their own "go-to-market engineers"), campus recruitment led by recent graduates, and go-to-market operations. One wheelie treats the sales team's automated meeting notes workflow as its own product, gathering feedback and iterating—support skills applied to internal customers.

Reversing Traditional Metrics

Clay spent a year trying to drive up first contact resolution, assuming it proxied for quality—close tickets fast with great first answers. Then they realized the metric created perverse incentives in an AI world. If a ticket can be resolved in one go, AI should handle it. Human agents should focus on consultative, multi-touch conversations.

So they reversed course entirely, now trying to drive first contact resolution down. Lower FCR means AI is successfully deflecting simple tickets, leaving humans for complex work that requires back-and-forth. This reversal required reframing the team's identity—not as efficient ticket processors, but as consultative problem-solvers for issues too nuanced for automation.

The shift extends to how Clay thinks about capacity planning. Traditional support teams project headcount scientifically: X tickets require Y people over Z months. The Wheel makes this impossible. Graduation dates vary, rotation timing shifts, and talent export timelines are unpredictable. Clay accepts this volatility as the price of the model, enabled by their small size and rapid growth creating constant rotation opportunities.

Junior PMs Embedded in Support

Clay positions support not as a cost center but as a value driver—the original mandate from co-founders down. This meant hiring people capable of doing more than answering tickets, then giving them the autonomy to act on what they learned.

One example: Tanvi (now a product marketing manager) noticed engineers weren't using Clay enough to understand customer issues. She created "engineering table builds"—sitting engineers down to build Clay tables as if they were customers, making visible all the friction points support saw daily. This wasn't assigned; she identified the gap and filled it.

Before The Wheel formalized rotations, everyone on the support team had "side projects" like this. AI deflection opened up capacity that naturally flowed to other work. Formalizing it just made the implicit explicit—support people should be spending significant time outside the queue, as long as quality metrics hold.

The defensive posture matters: Clay isn't cavalier about customer experience during this experimentation. They brought in Wendy from the people team to own rotation structure and setup after early cohorts hit bumps navigating ambiguous expectations. Dilthey explains, "We always decided that we really wanted support to be a value driver for the company. We really believed that if we had really good people on the ground talking to customers all day, every day, that would just pay dividends." The goal is rapid iteration with safety nets—move fast by preparing people carefully, not by accepting casualties.