# zapier-product-and-platform Published articles for zapier-product-and-platform. This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles. ## Which AI model is best for workflow automation? DevFeed: [Which AI model is best for workflow automation?]() Original publisher: [Read original article]() Author: Tracy Lee Published: 2026-10-28T07:00:00Z Content type: article Language: en Sources: [Zapier]() Topics: [workflow automation](), [Automation](), [AI Chat](), [Frontier AI]() Tags: [agent](), [ai](), [ai-agent](), [ai-model](), [ai-models](), [automation](), [automationbench](), [benchmarks](), [best](), [business](), [can](), [collection](), [compare](), [covering](), [fields](), [finance](), [healthcare](), [model](), [workflow](), [workflow-automation](), [zapier-product-and-platform]() ### AI overview AutomationBench measures end-to-end business workflows by checking the actual changes agents make across simulated apps, rather than their reported answers. Its 657 private tasks include ambiguous records and rules hidden in email; every check must pass. On the reported leaderboard, Gemini 4 Argon leads at 51.3%, so even the top model fails roughly half the time. Price does not reliably track performance: GPT-6 Sol scores about 80% of GPT-6 Astra at roughly one-sixth the cost per task, while Claude Sonnet 5.5 beats the more expensive Opus 5.5. GPT-6 Luna and DeepSeek V4 Flash cost $0.04 per task but complete fewer than a quarter of full workflows, making them better suited to individual steps. The source recommends stronger models for multi-app work, checking results rather than trusting agent reports, and adding a lookup and stop filter for workflows affecting customers or money. Scores reflect each model's best-scoring setting; Claude results use a fallback to another Claude model when a step is declined. ### Source excerpt Today, AutomationBench joins Fireworks' Specialized Intelligence Index (SII), bringing business workflow automation to a growing collection of practitioner-built benchmarks that compare AI models on real work. We're one of 32 new partners in this round, alongside teams covering fields like finance, healthcare, software, and ML research. AutomationBench measures whether an AI agent can finish a business workflow end to end across sales, marketing, operations, support, finance, and HR. Being part ## Use one Zapier MCP server across all your AI tools DevFeed: [Use one Zapier MCP server across all your AI tools]() Original publisher: [Read original article]() Author: Joe Stych Published: 2026-10-20T00:00:00Z Content type: tutorial Language: en Sources: [Zapier]() Topics: [MSP MCP](), [AI Chat](), [Tool]() Tags: [actions](), [ai](), [ai-tools](), [apps](), [chatgpt](), [claude](), [client](), [clients](), [connect](), [control](), [copy](), [cursor](), [mcp](), [mcp-server](), [server](), [set](), [share](), [skills](), [team](), [tools](), [zapier-product-and-platform]() ### AI overview Zapier MCP lets multiple AI tools share one server's apps, actions, skills, mode, and tool call history, so teams can configure them once. Server mode governs every connected client; managed mode lets the owner choose which actions clients can run, with account and workspace restrictions still applying. Each client connects to only one server at a time, and adding it to another moves it from its old server and adopts the new server's mode. Skills work only in agentic mode, so managed-mode servers share apps and actions but not skills. Zapier MCP is included in every Zapier plan, and Owners or Editors can add clients. ### Source excerpt Most teams don't stick to one AI tool. Your support team works in Claude, a few people swear by ChatGPT, and your engineers live in Cursor. Until now, each of those tools needed its own Zapier MCP server, with its own apps and skills to set up. Now you can connect all of them to one Zapier MCP server, and they share everything on it. Set up your apps and skills once Say your support team uses Claude and your engineers use Cursor, and both connect to the same server. Add Zendesk and its actions t ## Export your Zapier usage history as a CSV DevFeed: [Export your Zapier usage history as a CSV]() Original publisher: [Read original article]() Author: Joe Stych Published: 2026-10-15T00:00:00Z Content type: release Language: en Sources: [Zapier]() Topics: [CSV](), [SaaS Management]() Tags: [beta](), [billing](), [csv](), [export](), [history](), [zapier-product-and-platform]() ### AI overview Zapier's open beta export lets eligible accounts download task usage for up to 12 completed billing periods as a CSV, with one row per period showing dates, days, plan, task limit, total tasks, and tasks above the limit. Use it to compare usage against limits or chart changes over time; `days_in_usage` helps account for short periods after a plan change. The file excludes the current period and Free-plan periods, covers the whole account without Zap-level detail, and contains no prices. Counts use whole days in UTC and precede task credits, so they may differ slightly from invoices. Available on Free, Professional, and Team plans with monthly task limits, but not Enterprise or annual-limit plans; Team access is for Admins and Owners. Export via Billing & usage -> Usage insights -> Export usage history. ### Source excerpt Deciding whether your plan still fits means looking past the billing period you're in right now. The Usage insights tab shows the current period, but plan decisions usually need the last several. You can now export up to your last 12 completed billing periods of task usage as a CSV from the Usage insights tab. Each row puts the tasks you used next to your plan's task limit, so you can see at a glance which periods ran close. The export is in open beta. Right-size your plan Compare each period's ## Zapier Agents is now AI by Zapier DevFeed: [Zapier Agents is now AI by Zapier]() Original publisher: [Read original article]() Author: Ruchira Kulkarni Published: 2026-10-06T04:00:00Z Content type: tutorial Language: en Sources: [Zapier]() Topics: [AI Agent](), [Loop Engineering](), [agentic workflows](), [Multi-step AI processes](), [teamcity]() Tags: [agents](), [ai](), [ai-costs](), [deterministic-guardrails](), [tool-calling](), [zapier-product-and-platform]() ### AI overview Zapier has moved the capabilities of its standalone Agents product into an AI by Zapier step inside the Zap editor. The article explains how the integrated step can combine AI decisions with deterministic actions, require approval for sensitive tools, return structured output, detect needed tools, and show execution details in Zap History. It also describes task-based billing, per-run limits that pause runs above 75 tasks, and how to recreate eligible agents as AI by Zapier steps, with some capabilities still rolling out. ### Source excerpt Every capability that made Zapier Agents powerful--tool calling, reasoning, autonomous action--now lives inside a single AI step in the Zap editor, using AI by Zapier. Here, I'll map each Zapier Agents feature to its equivalent in AI by Zapier, so if you're coming from Zapier Agents, you know exactly where to click. What's changed between Zapier Agents and AI by Zapier Zapier Agents consisted of four things: a prompt, a set of tools, a trigger, and permission to act. That bundle worked, but it sat