The Real Lesson from Optimizely's Agentic AI Deployment

First, map the actual work flow before deploying any agent. Optimizely spent months understanding how work moved through its marketing organization. That upfront investment revealed where automation would help and where it would simply speed up waste.

Second, focus as much on workflow design as on agent capability. Production-grade adoption begins when work becomes repeatable and governable, not when the first agent goes live, Cicman said.

Third, governance and continuous improvement are not afterthoughts. As marketing became more systematized, Optimizely developed new approaches to supervision and iteration. Constraints surfaced, were resolved, then new ones appeared. "Managing constraints is a continuous process," Cicman said. "The most successful organizations build the ability to continuously find constraints, resolve them, and discover where they have moved next."

Technology provides the systems thinking. Marketing provides the customer understanding, creativity, and judgment. Agents provide execution capacity. The operating model turns all three into business results, Cicman argued. Technology leaders and marketing leaders must design these systems together. Neither group can do it alone.

For organizations deploying DXP AI agents, the takeaway is direct. Automation does not remove constraints. It reveals new ones. The future of agentic marketing will be shaped less by individual agents and more by an organization's ability to continuously redesign the systems in which those agents operate, Cicman said.

Forrester clients can read the full Customer Zero case study on Optimizely's AI governance approach.