How Zapier’s Teams Condensed Days of Work into Minutes with Claude Cowork

Zapier is the automation company connecting nearly 8,000 apps for more than 3.4 million businesses. AI adoption runs unusually deep: 97% of employees use AI in their daily work, thousands of agents are deployed internally, and Zapier was among Anthropic’s earliest MCP partners. When Claude Cowork launched, three teams picked it up and pushed it in distinct directions.

Even with high AI fluency, work requiring coordination across multiple systems remained bottlenecked. Product marketing homepage redesigns required design handoffs before anyone could evaluate a direction. Engineering analysis meant pulling data from six separate systems and stitching results manually. Influencer reporting lived in disparate tools with no unified view against quarterly goals.

Each team connected Cowork to Zapier’s full data infrastructure: a Databricks warehouse unifying 11 source systems including GitLab, Jira, Productboard, OpsLevel, Jellyfish, and incident tracking, plus org databases, Slack, and GitHub. Teams also built custom skills—reusable instructions encoding team-specific processes, voice, and positioning guidelines—so context accumulates across sessions rather than resetting each time.

AI Automation Engineer Larisa Cavallaro ran 15 SQL queries synthesizing live data from six engineering systems in a single session, producing an interactive dashboard that quantified inefficiencies and surfaced productivity drains the team hadn’t previously identified. Head of Product Marketing Joe Stych now delivers a CMO-ready homepage draft in about 15 minutes—work that previously required multi-team handoffs before anyone could evaluate direction. Senior Manager of Influencer Marketing Matt Brown built a daily-updated live dashboard on GitHub Pages tracking reach, clicks, and conversions against quarterly targets, turning retrospective reporting into forward-looking investment guidance.

The skills layer compounded the impact. Rather than re-prompting from scratch each session, context builds across sessions—every interaction makes the next one smarter. Zapier has since organized its internal AI architecture around Skills and MCP tools as the core reusable units, a framework that mirrors how it already thinks about app integrations. “The barrier between ‘having an idea’ and ‘shipping something’ has collapsed,” said Cavallaro. “I genuinely cannot remember doing my job without it.”