The Operations Confession

Zuckerberg told Meta employees on July 2 that AI agent development “has not accelerated in the way we expected.” Four months after cutting 10% of the workforce. After reassigning 7,000 people to AI teams. After raising capex guidance to $145 billion. The agents still don’t work.

Meta does not have a model problem. They have Llama, partnerships with Anthropic, compute that rivals nation-states. In January, executives were “super optimistic.” By July, the CEO is recalibrating timelines and admitting the reorganization “wasn’t as clean as it could have been.”

Same week, AWS announced a billion-dollar investment in forward-deployed AI engineers. Not selling models. Selling the operations layer — the people and process work that sits between a capable model and a useful outcome. The largest cloud company on earth looked at the market and concluded the gap is not technical.

This is the pattern. Companies assume spending solves it. Buy the tools, assign the team, announce the initiative. Then discover that nobody mapped how work actually flows. Nobody built the handoff protocols. Nobody defined what “done” looks like when a system does the work instead of a person.

The organizations pulling ahead treated AI as an operations problem from day one. Everyone else is still reorganizing.