The artificial intelligence industry has a glaring problem: 95 percent of companies experimenting with the technology are not making money from it, according to a Massachusetts Institute of Technology report published last month. Jason Droege, the new CEO of Scale AI, says the common assumption that organizations can simply plug in a model and see value is misguided. After a leadership shakeup tied to Meta acquiring a 49 percent stake in June, Droege argues that success depends on picking the right problems and building applications grounded in a company’s own data, not generic chatbots.
Scale AI is best known for data labeling that trains large language models, a service long used by major developers. Meta’s deal raised questions about whether rivals would keep working with Scale; OpenAI and Google have reportedly scaled back their engagements. Even so, Scale says its core data-labeling business has grown every month since the Meta investment. Droege, who joined last year as chief strategy officer, is doubling down on the company’s lesser-known applications arm, which helps clients assemble custom datasets and build tools that automate repetitive, error-prone work.
In Droege’s view, the companies that struggle are often applying Artificial Intelligence to the wrong problems or expecting a magic wand. He points to tasks where humans are slow or inconsistent, like reading and summarizing large volumes of documents, as better fits. Scale has supported systems that process insurance claims and prepare concise medical history summaries for doctors ahead of visits. Government agencies are also using Artificial Intelligence to pre-score building permit applications before human review. Crucially, Droege says, domain experts must be in the loop to provide feedback and improve outputs, especially in sensitive fields like healthcare. These projects can take weeks or months to stand up, but he contends they yield tools more useful to employees than off-the-shelf chatbots.
Scale’s client list spans the Mayo Clinic, the Qatari government, Cisco and Global Atlantic Financial Group. The company also recently signed a contract with the US Defense Department to develop Army applications. Droege says the applications business is already generating hundreds of millions of dollars in revenue, and that data operations have expanded monthly since the Meta deal. Still, analysts caution that broad, revenue-generating deployments could take years. Gil Luria of DA Davidson expects a long rollout cycle but believes organizational tools will eventually create significant value. With competition from Amazon, Microsoft and many others, Scale is pitching itself as a specialist that can help enterprises choose the right use cases. That stance aligns with MIT’s findings that organizations that tried to build Artificial Intelligence alone were the least successful.
