ChatGPT now raids your Slack and Google Drive for answers

OpenAI just gave ChatGPT a key to the kingdom of internal corporate data. The new ‘company knowledge’ feature, rolling out today for Business, Enterprise, and Edu users, lets the model pull context from connected apps like Slack, Google Drive, SharePoint, GitHub, and Intercom. The pitch is straightforward: stop hunting for information scattered across your tools and just ask a question. A version of GPT-5 trained to synthesize across multiple sources does the heavy lifting, and every response includes citations pointing to the original snippets and documents.

What makes this feel like more than a simple RAG (retrieval-augmented generation) wrapper is the reasoning layer. The system can run multiple searches to untangle conflicting details. If your team hasn’t aligned on next year’s goals, ChatGPT won’t hallucinate a consensus — it will surface the differing viewpoints. It can also handle time-sensitive queries by ranking sources by recency and quality, pulling Q1 performance data without forcing you to dig through old files. The sidebar lets you watch it search in real time, which is a nice touch for trust but also feels like a flex.

Admins get a full control panel. Enterprise and Edu accounts can set group-level permissions, create custom roles, and use the existing Enterprise Compliance API for conversation logs. OpenAI is firm on the privacy line: the model respects existing file permissions by default and doesn’t train on your data. SSO, SCIM, and IP allowlisting are all baked in. That’s table stakes for enterprise SaaS in 2025, but it’s worth noting that the permissions model only works if companies have their internal access controls in order — a big ‘if.’

The real test will be noise. Giving an LLM access to Slack messages, email threads, and support tickets sounds powerful until you realize how much of that content is half-baked thinking, outdated decisions, and random emoji reactions. OpenAI frames this as a way to get a ‘more complete picture,’ but completeness isn’t always clarity. Sometimes you just get a more complete mess. Still, for teams already drowning in connected tools, the promise of making that sprawl coherent — however imperfectly — will be hard to resist.