We took a national creative studio’s capacity within a single campaign from about forty assets to over 700.

Trade School is a creative production studio within Chalkboard Ventures that builds campaign work for national advertisers, and the pressures its clients brought to the studio had become common across the category: fragmenting channels, expectations of relevance at the audience level, compressed time to market, and a heavier burden to prove performance on every placement. Satisfying those expectations required more creative, in more variants, for more platforms, on shorter notice than the studio’s production model could absorb manually.

Production effort scaled linearly with volume, because each additional audience, funnel stage, or aspect ratio demanded another cycle of manual resizing, rewriting, and quality control, which held a typical campaign to roughly forty assets and a handful of audience segments. Platform vendors and holding-company production arms were pitching creative automation directly to the accounts Trade School was defending, which turned a production challenge into a competitive one, and the studio came to Eskridge to build the capability that would let it scale without giving up its creative standards.

Creative automation platforms were already fast, and the constraint they shared was that their output arrived finished and unalterable, which was the wrong trade for a studio whose clients had grown accustomed to the flexibility of bespoke work and expected to give notes on a headline or a crop the way they always had. We made the product decision that shaped everything after it, generating assets directly into Figma instead of a closed rendering engine, so that Claude could write the copy for each unit and generative outfill could extend approved photography across every aspect ratio a media plan called for while the design team kept the ability to open any frame and change what came back.

Once AI was carrying the copy and the imagery, the bottleneck moved upstream into campaign setup, where a producer was still transcribing audiences, placements, funnel stages, and messaging direction out of briefs and media matrices into Airtable by hand before anything could be generated at all. We turned to Claude again and built agents to ingest those documents directly, so that the brief and media plan a team has already written became the system’s inputs, which also addressed a quality problem from the first pilot, where copy generated against campaign-level context alone came back generic at the product level. We have since partnered with Trade School to pitch Mainspring as a managed service, where the senior creative talent running the system is as much a part of what a client buys as the technology underneath it.

The first production campaign ran three Oldcastle APG brands through the system and delivered 716 traffic-ready assets across paid, social, and video in 47 days from creative brief to final delivery, an eighteenfold increase in creative production throughput against the studio’s previous ceiling of roughly forty assets per campaign. We also reduced Trade School’s internal working days by roughly a third, which compressed a comparable engagement from thirty days to twenty-one while preserving client review time in full.

Our engagement has since widened from an engineering build into an agency-wide AI program across both Chalkboard agencies, covering an enterprise Claude deployment, an OKR framework for tracking adoption and impact, and a standing fractional technology leadership role.