Romiel InolinoRomiel InolinoAI & CRM Automation Specialist
All posts
2 min readby Romiel Inolino

ChatGPT plug-ins can now trigger automations from app events. What DevDay means for workflow builders

OpenAI DevDayMCPautomationn8nZapier

Event triggers, fast routing decisions and agents that operate software. OpenAI just shipped the three building blocks of a workflow tool.

What happened

At DevDay, OpenAI expanded ChatGPT plug-ins, which already connect to tools like Slack, SharePoint, Airtable and Google Drive. Plug-in extensions get a dedicated sidebar home, interactive panels and file viewers, and users can approve a plug-in's access needs individually, per TechCrunch. OpenAI also said it is adding support for the proposed MCP Events specification, which would let plug-ins start automations based on events in a connected app, and making automations easier to add and manage.

On the API side, The Decoder reports a new Decisions API built on GPT-6 Luna for fast, single decisions. Developers define a question with a limited set of predefined answers, pass in context, and get an answer back, for classifying content, routing requests or picking an agent's next step. OpenAI says it decides ten times faster than GPT-6 Luna through the regular API. It launches as a limited preview. The Agents API, in public beta since September 11, also gains Computer Use.

My take

Put those together: an event arrives, a cheap fast model classifies it, an agent acts, including in software without an API. That is the shape of most automations I build in n8n, Make or Zapier.

I do not think this replaces those tools yet. The Decisions API is a preview and MCP Events is still a proposed spec. But it changes a few design choices now:

  1. Routing steps should be classifiers, not chat. If a step only picks from "sales, support, billing, spam", constrain it to a fixed answer set. It is cheaper, faster and easier to test.
  2. Build on MCP where you can. If triggers and actions both speak MCP, your workflows get more portable between platforms.
  3. Treat Computer Use as a last resort. When a system has an API or webhook, use it. Screen control is for the legacy tool with no other way in.
  4. Keep the audit trail in one place. Automations spread across ChatGPT, a workflow tool and a CRM get hard to debug fast.

Designing the flow so each step uses the simplest reliable tool is the kind of system I build for teams. romielwillautomate.dev

Sources

More posts