Wolff & Park, a 12-person agency in Chicago, ran a 90-day experiment in mid-2024: their social content team continued producing content for half their clients as usual, while AI tools (Claude for copy, Midjourney for imagery, custom workflows for scheduling and adaptation) handled the other half. The results are instructive.
What AI Did Well
Volume, consistency, and response time. The AI workflow produced content at 3x the volume, maintained consistent brand voice across formats, and could respond to breaking news or trending topics within 45 minutes instead of the usual 3-hour production cycle. For commodity social content — announcements, promotional posts, evergreen educational content — performance metrics were essentially identical to human-produced content.
Where Human Work Outperformed
Cultural moment response, humor, and anything requiring subtle audience understanding. The AI consistently produced content that was technically correct but slightly off in tone for real-time cultural moments — the kind of slight wrongness that audiences feel before they can articulate it. Engagement rates for human-created reactive content outperformed AI versions by 34%.
The Business Decision
Wolff & Park now uses AI for 70% of social content production, with human creative direction on all output and human-only production for cultural moment content. The efficiency gain funded three new strategy roles that had been backlogged for hiring. Net effect: better-staffed strategy capability, maintained creative quality, 40% lower production costs.