Meet the Data Agent in ChatGPT Work

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: Meet the Data Agent in ChatGPT Work · Channel: OpenAI · 03:11 · Watch on YouTube

Topics: AI agents, Data, databases and embeddings, Prompt and context engineering

Summary

The video shows how the ChatGPT work data agent uses trusted company data and existing business context to turn a question about a product launch into actionable recommendations. It helps investigate customer behavior, review the evidence behind an analysis, assign next steps, and share progress through a dashboard and alerts. The data team can create a context layer from warehouse data, conversations, and documentation so people across the organization use consistent definitions and get trusted results.

Key points

A manager asks about the progress of a new product launch from a phone, and the data agent analyzes trusted data using existing site context.
The analysis finds that returning creators are driving growth, while only about 16% of people with access have started creating.
The data plugin lets users inspect metric definitions, underlying data, and SQL queries, while the data team can review how answers were produced.
Slack conversations and Google Drive reports suggest customers are unsure where to start.
The proposed action is to give users an in-app pointer on how to publish a site, then assign owners and next steps.
The data plugin can turn the analysis into a dashboard that teams use to explore adoption, returning creators, and customer support needs.
Users can define an alert for a week-over-week drop in creator adoption, including its threshold, timing, and recipients.
A data context layer combines warehouse data, analytical conversations, and documentation to define business terms and consistent comparisons.

Chapters

Start	Chapter	Summary
00:01	From a business question to launch analysis	The data agent investigates a product launch question using trusted data and reveals that relatively few people with access have started creating.
01:02	Recommendations and shared action	Feedback from Slack and Google Drive points to uncertainty about where to start, leading to an in-app recommendation and assigned next steps.
01:35	Dashboards and adoption alerts	The analysis becomes a shared dashboard, and the team can define an alert for a drop in weekly creator adoption.
02:39	Creating consistent data context	The data team builds a context layer that records business definitions and helps ensure consistent, trusted answers across the organization.

Quotes

Your company has invested in data. Now let’s put it to work. — 00:01

Only about 16% of people with access have started creating. — 00:32

The feedback points to uncertainty about where to start. — 01:02

This context spells out what our business terms mean. — 02:39

they’re working from the same business context and getting consistent trusted results. — 03:11

Concepts and entities

ChatGPT Work, data agent, data plugin, Slack, Google Drive, SQL, data context layer, data team

Full transcript (original language)

00:01 Your company has invested in data. Now let’s put it to work. The ChatGPT work data agent brings trusted data and your data team’s expertise into everyday decisions. Let me show you how we can go from a business question to a plan the team can act on. On my way into work, my management team asks how our new product launch is going. I send the question from my phone. The data agents get to work using our trusted data and existing sites context.

00:32 By the time I’m at my desk, the analysis is ready. Here’s the opportunity. Returning creators are driving growth, but look at the funnel. Only about 16% of people with access have started creating. We need to help more customers take that first step. With the data plugin, I can dig into evidence, the metric definitions, the underlying data, and the SQL queries behind the report. Our data team can inspect how the answer was produced and check that it answers the right business questions. Now I can check Slack conversations and existing reports in

01:02 Google Drive for clues about why customers aren’t creating sites. The feedback points to uncertainty about where to start. That gives me a practical recommendation. Give users an in-app pointer on how to publish a site. With those recommendations, I ask ChatGPT to help assign owners and next steps. I review the plan and share it in Slack. Now the right people can take action with clear responsibilities and a shared way to measure progress. Next, let’s give the team a shared view of progress. The data plugin turns this

01:35 analysis into a dashboard, so product leaders, customer success, and the data team can work from the same trusted sources and definitions. Now everyone has a place to answer the questions that come up every week. Are more customers adopting this product? Are creators coming back? Which customer groups need more help? Teams can explore those questions themselves, while the data team focuses on deeper investigation. I also want to be alerted if weekly creator adoption drops compared with the previous week. I asked ChatGPT to help define the rule, how large a drop matters, and when to check, and who

02:07 should hear about it. Review those details before scheduling. With refresh options, we can keep that shared view current throughout the launch. The value is an ongoing way to spot changes, agree where to focus, and decide when to adjust the plan. And you’re probably wondering how we can create consistent data across the organization, and that’s where context comes in. The analysis and dashboard we just saw use an existing site’s context layer. Now, let’s look at how the data team created that foundation so future

02:39 questions use the same business definitions. Here, I used the data agent to help create a data context layer from our warehouse, analytical conversations, and documentation. It brings together the knowledge our data team uses to interpret the data. This context spells out what our business terms mean. A person who views a site is different from someone who creates one. It also captures which sources and activity count, what reporting period to use, and how to make consistent comparisons. Now, I’ve got consistent metric definitions across my organization. So, when anyone asks

03:11 questions of the data, they’re working from the same business context and getting consistent trusted results. The data plugin brings trusted analytics, shared dashboards, and consistent business context into everyday work. Your data team’s expertise can now support everyone who needs to make a decision. Now, you can put trusted data to work wherever decisions are made.