AI Pulse · Business Models & Winners

Sierra just put a long-running AI agent into production for BBVA in 30 days, according to the vendor’s announcement. The agent is live in Spain and Argentina. BBVA calls it their first “Horizon” agent, whatever internal label that carries for them.

Thirty days is the number that matters here. Not a pilot. Not a proof of concept that sits in a sandbox for six months while committees debate governance. A named bank, two countries, production traffic. Sierra did not release deal size or usage figures, so treat this as a directional signal rather than a full case study. But the shape of the story is what enterprises shopping for agent platforms will now expect: a hard timeline, a real deployment, a brand willing to go on record.

This is the kind of reference point that moves budget conversations. If your leadership is asking why your AI projects take quarters instead of weeks, they will hear about examples like this. The gap is not always capability. It is often process, vendor readiness, and whether the use case was scoped tightly enough to ship fast. A 30-day sprint implies BBVA knew exactly what problem the agent would solve before they started, and Sierra had the rails in place to move quickly once the decision was made.

If you are building something similar internally, the lesson is not that you need Sierra. It is that speed to production is now a competitive benchmark. The tools exist. The question is whether your workflow can support that pace. This is the kind of thing we build into an AI command centre at Omni, where the goal is not just deploying agents but doing it repeatably without reinventing the stack each time.