OpenAI Data agent in ChatGPT Work connects warehouses and BI

Generated by Grok Bot research on 2026-09-10. WebFetch + native X. Treat as raw material — review before promoting into a project or thread.

Summary

On 2026-09-10 OpenAI announced a new Data agent inside ChatGPT Work, available as a Data plugin: connect approved company data sources, ask plain-language questions, investigate changes, and build shareable interactive dashboards without writing SQL. Official connectors named include Amazon Redshift, Datadog, Google BigQuery, ClickHouse, Databricks, MongoDB, Snowflake, plus Google Drive and SharePoint files; BI surfaces include Omni, Oracle BI, Power BI, Sigma, Tableau, and ThoughtSpot. VentureBeat reports OpenAI has not published an external retrieval/accuracy benchmark for the product (internal parity checks only). The primary @ChatGPT announcement post includes video — run /transcribe-clipping before promote.

Findings

Product launch (issuer)

OpenAI’s Now everyone can put data to work (dated September 10, 2026) introduces a Data agent in ChatGPT Work that “connects to your company data, investigates what changed, and builds interactive dashboards you can share,” directed in one conversation without writing queries.

Named data / platform connections on the issuer page: Amazon Redshift, Datadog, Google BigQuery, ClickHouse, Databricks, MongoDB, Snowflake, “and more”; files from Google Drive and SharePoint. Semantic/context partners cited include Databricks Genie Ontology, dbt, GitHub, Snowflake Horizon, and BI dashboards.

Named BI build/interact targets: Omni, Oracle BI, Power BI, Sigma, Tableau, and ThoughtSpot.

Admin / permissions (issuer claim): Enterprise admins choose which connections and roles are available; “Queries enforce the connected account’s existing permissions, including table, row, and column restrictions.”

Install path (issuer): Listed as Data in the Plugins directory in ChatGPT Work; admins can make it available or install via Workspace settings → Plugins; users start with @Data.

Internal adoption (issuer claim, not independently verified): “Nearly all of our product team and over two-thirds of our GTM organization use data agents in ChatGPT Work.” Alpha / customer quotes on the same page include NTT DATA, Thermo Fisher, ServiceTitan, Zipline, Empower, and others — treat as marketing testimonials, not measured outcomes.

Positioning and contested gaps (secondary)

VentureBeat (Sean Michael Kerner, 2026-09-10) frames the launch as externalizing an internal OpenAI data agent, quotes GM Arpan Shah on prior low self-serve answer rates, and states OpenAI did not publish a retrieval accuracy / correctness benchmark for the external Data agent (internal comparison only). VB also quotes Shah that the agent stitches partner contexts “in their state” rather than building a persistent customer context layer. Competitive framing vs Anthropic Cowork and Databricks instructed-retriever claims is VB’s analysis — keep as discourse, not issuer fact.

Native X

The product account @ChatGPT (2026-09-10 ~15:05 UTC) announced the Data agent with the same issuer URL and an attached video. Discourse reposts (e.g. connector lists) are secondary to the issuer post and blog.

Contradictions and open questions

• No published external accuracy / faithfulness benchmark for the Data agent (VB); issuer page emphasizes connectors and workflow, not measured answer quality.
• Does “ChatGPT Work Data agent” meaningfully outcompete warehouse-native agents (Databricks Genie, Snowflake CoCo, etc.), or mainly orchestrate across them? Issuer partners with those vendors — competitive boundary unclear.
• Video on the primary X post is not yet transcribed; do not treat demo visuals as verified capability until /transcribe-clipping.
• Unrelated same-day X chatter about an “Amodei Pentagon briefing” on DeepSeek is Community-Noted as unsubstantiated — not filed here; DeepSeek-V4.1-Flash already ingested in the 10am pass.

Provenance

Method: Grok Bot / WebSearch-equivalent news pointer / WebFetch / native X (search_news, get_posts_by_ids, get_users_posts) Generated: 2026-09-10

Web sources:

• Now everyone can put data to work | OpenAI — issuer launch post: connectors, BI targets, admin permissions, install path, internal/customer quotes
• OpenAI's new data agent skips … a benchmark | VentureBeat — secondary: Shah interview, no published benchmark, no persistent context layer claim

X sources:

• X post by @ChatGPT (2026-09-10) — official announcement + video (x_video) linking issuer URL
• Discourse cluster posts hydrated for context only; cite issuer permalink above as primary

Grokipedia:

• not used

Video transcripts

https://x.com/ChatGPT/status/2098065296968011853

Transcribed 2026-09-10 with Whisper base. Duration: 1m 11s. Language: English. Uploader: @ChatGPT.

awia Uık U deeply deeply deeply deeply deeply deeply deeply