Aug 2025 · Agentic AI · Automation · Creative Operations
Fixing the feedback loop.
The problem
Creative feedback looks simple on paper. Work gets submitted, someone senior reviews it, notes come back. In practice, the approver is busy, the calendar is full, and the work was due yesterday. A quick comment becomes a meeting, the meeting gets rescheduled, and by the time useful feedback lands it’s either too late to act on or the work has gone too far in the wrong direction. Four rounds and three meetings to get a sign-off that should have taken twenty minutes. Every creative team knows this problem. Most just accept it. The insight, after working with design teams, is that a lot of that precious 30 minutes with a senior stakeholder gets wasted on small, simple mistakes, distractions that eat into the real, valuable feedback the designer actually needs.
The approach
For my final project in Relevance AI's Ops bootcamp, I built an AI agent to support the feedback loop without replacing anyone in it. The agent learns how each approver likes to review and make decisions. For this build, I created a user manual for the Creative Director covering how they like to communicate and how they like creative work presented. It checks submissions against brand guidelines before anything reaches a stakeholder, catching simple mistakes early, and gives structured feedback using the 10/50/99 framework, calibrated to the stage of the work. Everything surfaces directly in Slack, so it's async, fast, and visible to the whole team.
The result
Work quality improves before it ever reaches review. Issues get caught early instead of surfacing in a meeting, and leaders spend their time on what actually requires their judgment, not correcting avoidable mistakes. The meeting doesn't start the feedback loop. It finishes it.
If you'd like to see this in action, get in touch and I'll send you the agent.