The Boring Part

OpenAI shipped Presence on July 22. Anthropic shipped Ode a week earlier. Neither product is a model. Neither is an API. Both are teams of engineers who fly to your office and wire AI into the workflows your people actually run.

Two competitors who agree on almost nothing arrived at the same structural answer within eight days. The bottleneck was never intelligence. It was permissions, policies, escalation rules, and the dozen legacy systems that touch every customer request.

Cisco says 85 percent of enterprises are piloting agents. Five percent have shipped one to production. That 80-point gap is not a technology problem. Nobody is stuck because the model cannot reason well enough. They are stuck because nobody on the team owns the deployment — the workflow redesign, the monitoring, the update process when a policy changes.

OpenAI tested Presence on its own support line first. Seventy-five percent of issues resolved without a human. Company-reported numbers, not audited, but the point stands: they needed their own forward-deployed engineers to make their own model work in their own company.

If the people who built the model need a consulting arm to deploy it, the rest of us were never going to solve this with a better prompt. The work is mapping the process, defining what the agent can touch, and watching what happens when it does. Not exciting. Not optional.