Jason Droege has a simple message for his 1,300 employees at Scale AI: stop chasing shiny objects and focus on what actually works in the real world. The newly minted CEO, who took over last June when founder Alexandr Wang departed to lead Meta's AI strategy, is pushing hard on a concept he calls "The Reliability Race" in a memo being circulated internally for the first time.
The stakes are real. Mission-critical operations in government, the military, and major corporations cannot afford AI systems that work 95 percent of the time. When an autonomous system makes a catastrophic error, the price tag in money, time, and lives can be enormous. That gap between promise and performance has become the central tension in enterprise AI adoption, and Droege sees it as Scale's competitive advantage.
"The cost of mistakes in these environments can be high," Droege said in an interview from San Francisco, where Scale marked its 10th anniversary this week. The 47-year-old executive spent time talking with customers and prospects before crafting his leadership mandate. What he kept hearing was not enthusiasm but frustration. AI products would run flawlessly most of the time, then fail spectacularly on edge cases. Executives knew the technology could work, but they could not justify the investment until reliability improved.
"We said: OK, this is an opportunity to bridge this gap between expectations and reality," Droege explained. "And so, that's what we've been investing in for the past few years."
His memo hammers home a point that runs counter to much of the AI industry's current messaging. "Reliability at this level depends on human intelligence," Droege wrote. Scale maintains what he calls "forward-deployed engineers" who work inside customer environments to tune and test AI systems for specific workflows. This is not sexy venture-capital material, and it contradicts the narrative of self-improving algorithms that need little human oversight.
The move also serves a strategic purpose. Scale has long pitched itself as a data annotation firm, the unglamorous back-office operation that labels images and text to train AI models. But Droege wants the world to see something different: a company that handles the messy, unglamorous work of making AI systems actually function in production. Last June, when Wang transitioned to Meta and the social media giant took a 49 percent stake in Scale, the company signaled its ambition to be more than a vendor.
Droege acknowledges the noise problem. "People are talking about hundreds of topics around AI constantly," he told Axios. "What I'm trying to do with the memo is really focus our team on the most important thing our customers tell us." He learned early in his career that leaders must "keep the main thing front and center constantly" while diving deep into the specifics. Distraction is a path to failure, especially in a field moving as fast as AI.
The reliability message also cuts through industry hype. "What we hear in the industry is lots of hype, lots of talk. Then Scale will innovate and grind to get to the outcomes that the customers expect," Droege said. He is betting that while others argue about AI's potential, Scale can build the infrastructure and expertise that turns potential into reliable dollars and cents.
Author James Rodriguez: "Droege's pivot to reliability is the right move at the right time, though it requires Scale to compete on execution, not just innovation, which is a harder sell to investors."
Comments