OpenAI has published two guides documenting how data science and sales teams use ChatGPT Work in practice. The guides showcase complete workflows: data scientists build root-cause analysis briefs, impact readouts, KPI memos, scoped analyses, and dashboard specs starting from real work inputs. Sales teams produce pipeline briefs, meeting prep packages, forecast reviews, account plans, and stalled-deal diagnoses.

Why this matters. As we reported on July 10, ChatGPT Work has moved from announcement to operational tool. These guides represent real adoption documentation: concrete use cases instead of generic demos, with the inputs these teams actually have and the outputs they need to produce. If your work resembles one of these scenarios, you have a verifiable roadmap to start from.

Where to look. The guides live in OpenAI’s Academy section. Each workflow is described with sample inputs, steps, and expected output. These aren’t tutorials to follow literally, but frameworks to adapt to your context: your data scientists work with the same materials (data, dashboards, briefs), and your sales team prepares the same things (pipelines, forecasts, diagnoses). The value lies in seeing the structure: what to give the agent, how to break it down, what to verify.

In detail

The context

ChatGPT Work was announced on July 10 alongside GPT-5.6 as an agent capable of operating on apps and files for extended projects. Until now, documentation on how teams were actually using it was missing: these two documents fill that gap.

What the guides say

The data science guide covers five recurring output types: root-cause analysis briefs (finding the cause of an anomaly), impact readouts (quantifying the effect of a change), KPI memos (summarizing an indicator’s performance), scoped analyses (answering a defined question with the right data), and dashboard specs (translating a visualization request into technical requirements). Each type starts from real materials: datasets, SQL queries, meeting notes.

The sales guide documents five workflows: pipeline briefs (current state of opportunities), meeting prep packets (materials for a customer meeting), forecast reviews (closure predictions with confidence levels), account plans (strategy for a specific customer), and stalled-deal diagnosis (why an opportunity isn’t closing). Again, concrete inputs: CRM, past emails, call transcripts.

What changes for ChatGPT Work users

Before these guides, the only available documentation was the general announcement. Now those in these roles have a verifiable starting structure: not a tutorial to copy, but examples of how other teams have broken down their recurring tasks. The structure matters more than the details: sample inputs, intermediate steps, expected output, what to verify.

These guides also show the constraints: each workflow assumes data is accessible and well-formatted, the agent has the right permissions, and someone verifies the output before use. It’s not complete automation—it’s assistance on tasks that still require oversight.

Implications

Publishing operational guides instead of just generic demos signals that OpenAI sees ChatGPT Work as mature for enterprise adoption. The guides are public: anyone can read and adapt them, no special business accounts needed. This lowers the adoption barrier, but also raises a training question: teams will need to understand how to adapt these workflows to their own processes without blindly copying them.

Another signal: the guides focus on intermediate outputs (briefs, memos, specs), not final decisions. The agent prepares the material, but the decision stays human. It’s consistent with the site’s angle: AI empowers people, it doesn’t replace them.