Enterprise

ChatGPT Work adds Gmail and Slack webhook triggers, moving inbound-lead response from scheduled sweeps to real time

OpenAI's August 25 update lets Plus and Pro users fire AI-drafted responses the instant a Gmail message lands or a Slack channel gets a post — and lets one person's task setup be shared across a team.

OpenAI extended ChatGPT Work with event-driven webhook triggers on August 25, replacing the scheduled inbox sweeps that had defined the product since its July launch as a modified version of the Codex coding tool. A qualifying Gmail message, a post in a monitored Slack channel, or activity on a GitHub pull request now fires a task the moment it arrives, instead of waiting for the next interval check.

The architectural distinction matters more than it sounds. Under the old model, OpenAI’s infrastructure checked connected apps at defined intervals, retrieved anything new, and ran the configured task. Under the new one, connected apps push HTTP POST notifications to an OpenAI-registered endpoint. Research cited by TechTimes put event-driven systems at 62% lower latency than polling equivalents. For a small B2B team without an SDR, that’s the difference between a prospect reply that’s answered inside the minute and one that goes cold in a queue.

Gmail triggers support sender and subject filtering. Slack requires you to “add @ChatGPT to each monitored Slack channel.” OpenAI’s release notes describe use cases like “summarize an email announcing that registration has opened, or draft next steps after a client shares feedback in Slack.” Shareable task templates arrive in the same update, so “someone else can review and customize a task, connect their own apps, and schedule an independent copy.” The Android build requires version 1.2026.216 or later.

Adoption is the awkward subplot. ChatGPT Work runs $20 a month, and an OpenAI-backed June study found 98% of OpenAI employees using Codex, versus 17% of organizational subscribers and less than 1% of individual subscribers. The combined Codex/Work app has roughly 20 million users against more than a billion for ChatGPT overall. TechCrunch describes the pitch as aimed at “white-collar workers to field AI agents — hooking LLMs up to the digital workflows used by accountants, investors, doctors, and everyone else whose day-to-day is dominated by their computer.”

The event-driven turn lands into a broader agent buildout, including the Salesforce–Anthropic integration putting CRM data inside Claude and Agentforce’s $1.5 billion ARR run on 3.2 billion completed tasks. It also arrives with real seams. Approval-gated actions pause; read, summarize, and draft actions don’t, and an always-on OAuth listener expands the surface for prompt-injection attacks embedded in incoming mail. OpenAI has said as much: “prompt injection, much like scams and social engineering on the web, is unlikely to ever be fully ‘solved.’” The EU AI Act’s Article 50 transparency obligations for agentic systems took effect August 2, three weeks before shipping. Free accounts get scheduled tasks for the first time, capped at 3 active tasks and no more than once per day, with no webhook triggers.

The shift from a 9 AM inbox digest to a listener that never sleeps is a small feature note and a large change in what an AI subscription is for.

Sources