Enterprises Turn Inward for AI Expertise

Businesses are increasingly opting to build their own specialized artificial intelligence models rather than relying on generic, rented solutions. This shift reflects a growing recognition that using the same foundation model as competitors creates no competitive advantage and potentially exposes proprietary data to outside vendors.

The Case for Custom AI

Enterprises are discovering that general-purpose AI models can be inefficient—costing 5 to 10 times more than specialized alternatives. Moreover, every query sent to a third-party model provides the vendor with valuable business insights. As Oumi CEO Manos Koukoumidis puts it, “Nearly all companies are running the exact same closed, generalized models trained on the public web.”

Oumi, a Seattle startup founded by ex-Google, Microsoft, and Apple engineers, launched its Compounding AI Factory to automate this process. The platform enables businesses to deploy customized models that continuously learn from their own data.

Early Success Stories

Several companies have already demonstrated the benefits of in-house AI:

• Morgan Stanley used an internal tool called DevGen.AI to translate 9 million lines of legacy code, saving developers an estimated 280,000 hours
• A top-five U.S. bank leveraged Oumi’s platform to modernize 100 million lines of code after a general AI failed initial tests
• Mistral has built a $400+ million business helping companies train AI on their own data, with clients including HSBC and Stellantis

Challenges Remain

While financial services firms have shown early leadership in this trend—with high adoption across 27 of 38 tracked tasks—data fragmentation remains a significant barrier. According to PYMNTS Intelligence, 30% of financial executives cite data quality as their biggest deployment challenge.

This underscores the fundamental truth that owning specialized AI requires organized, accessible data—a capability many companies currently lack.