Zapier

This photo was taken at an eaRISE + eaMAFIA Hackathon, and it pretty much sums up the last few weeks. I've been building an AI operating model for our exec team. Shared brains with instruction layers, automated meeting and coaching pipelines, interactive apps, AI ops playbooks, etc. It's the kind of work that's hard to explain. But now people have heard about how it's running and asking to understand it. ‣ With eaRISE, 100 or so folks built practical dashboards with Zapier MCP and their agent harness. ‣ At Zapier's Outpost in Cape Cod, I facilitated alongside execs leading AI transformation at their own companies. They wanted to see my shared brain, the apps I'm building, how I set it all up. ‣ Back at work, an engineering zone lead DM'd asking for a full walkthrough of our exec brain. ‣ Marketing wants to turn one or more of my playbooks into a ZapConnect session. I'm watching what I build and share ripple to other teams & companies 🤯 I came up through public relations, a startup distillery, then EA work. But somewhere along the way I'm now showing how an exec and EA team operates in an AI world. I'm still not sure I fully believe it! But I believe in the work, and the people I'm doing it with. Thanks to the people who've trusted me enough to ask questions, share what they're figuring out, and let me be part of their work.

People have already run 4.6M+ tool calls through Zapier MCP. Here's what they're actually building: • an AI agent with its own email address that runs an entire brokerage • a Customer Support ticket system that saved 917+ hours/month • custom interview briefing docs pulling from 4+ connected HR tools • weekly goals review that searches Slack and email for evidence of what you actually accomplished • morning dev report that pulls MR status, reviewer comments, and priority actions from GitLab • Landing pages built from scattered Google Docs (drafted, SEO'd, and pushed to the CMS) • Oh, and some hyper-specific Spotify playlists Think you can top those? Try it out here: https://zapier.com/mcp

Your wish is our command: the "You Can Call Me Al" dance tutorial from the choreographer herself... so you can dance just as well as you automate 🤝

Anthropic launched Opus 4.8 today. They benchmarked it against AutomationBench, built by our team here at Zapier. Here's what AutomationBench actually does: it drops an AI agent into a simulated company running 47 apps and dozens of REST endpoints, gives it a single natural-language instruction, and watches whether it can complete a real business workflow. No handholding. The agent has to discover the right APIs, chain interdependent calls, follow layered business rules, and avoid deliberately planted distractors. Pass/fail, deterministic grading. Opus 4.8 scored 15.5% on our set, up from 9.9% on Opus 4.7 (!!) Proud of the team that built this. Big shoutout to Daniel Shepard, Robin Salimans, Lukas Bergstrom, Michael Haarala and everyone And proud that when Anthropic wanted to show what their newest model could do with real tools and real workflows, they came to us. Here's why that matters beyond the benchmark. Our MCP and SDK customers are doing this exact thing every day. Giving an agent a goal and letting it orchestrate across their apps. Models are getting better fast! The bar is rising with them. #Zapier #AI #MCP

Opus 4.8, the first model to break 15% on AutomationBench, is now live in Zapier! It handles complex HR, Finance, and multi-app workflows better than anything else we've tested: refusals dropped from 20% to 4% Opus 4.7 would see a sensitive task and stop, but 4.8 keeps going AutomationBench results in Anthropic's Opus 4.8 system card

AL is back: this time, in a music video Zapier's "You Can Call Me AL" is live!