Off-the-shelf software gives everyone the same capabilities, which means it gives no one an edge. The companies pulling ahead are building thin, sharp internal tools on top of frontier AI models — tuned to their content, their data and their workflow.
The hard part isn't the model. It's knowing which workflow is worth automating, what the output has to look like before an editor trusts it, and where a human stays in the loop. That is operator knowledge, not engineering knowledge — and it's where most AI projects fail.
I've spent 25 years inside these workflows and the last several building the tools: content intelligence systems, search prediction engines, editorial databases and video tooling. I scope hard, ship a working prototype fast, and harden only what earns its place.