In 35 years in this business, I have never seen anything quite like the present moment. Industries are moving at once—and betting on the whole racetrack.

Earlier this year, I sat with a client whose proof-of-concept for a single, narrow-agent workflow involved 14 different vendors stitched together. He was rather proud of it. I asked him, gently, how he planned to put it into production. He looked at me as if the question hadn't occurred to him, and said, "We haven't really got to that part yet." I showed him a competitor who had and was already live in production after eight weeks. He stopped talking.

That small private moment is, to my mind, the defining moment of agentic AI in 2026: A smart team realizing they have built a beautiful demo that can't be turned into a business.

I’ll close with a moment that has stayed with me. On the day his team went live, the CTO of a very large retailer told me the project we had built together was obsolete. I was, for a moment, heartbroken. Then he explained, “Boris, it’s obsolete because it’s live,” he said. “But the data was MongoDB. That is my foundation. I can replace the LLM and the model at any time. I can replace the framework on top. The data is mine.”

What we’ve learned from making these big moves is that the architectural choices made now will outlast the AI tools they were chosen for. That, in the end, is why the data layer is the decision worth getting right.