I'm a founder and decadelong Microsoft engineer. Here's who I'd hire to lead the next stage of AI.

This opinion piece argues that most companies are approaching AI adoption incorrectly by focusing too much on buying off-the-shelf AI software or waiting to build proprietary models. The author, a former Microsoft engineering leader, says the real near-term opportunity is customizing existing large language models to fit company workflows, data, and institutional knowledge. He highlights a practical rollout strategy: start with small, business-close wins rather than a top-down AI mandate. The article also identifies a key talent pool for leading AI transformation: not just AI researchers or developers, but “crafters” — employees who already build spreadsheets, dashboards, scripts, and automations to solve operational problems. The market implication is that enterprise AI value will likely accrue to firms that integrate AI into workflows effectively and to software/data tools that enable this customization, rather than to generic AI products alone.