Use ChatGPT Work to ground analysis in your semantic layer

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: Use ChatGPT Work to ground analysis in your semantic layer · Channel: OpenAI · 01:35 · Watch on YouTube

Topics: AI agents, Data, databases and embeddings

Summary

Trustworthy data analysis with ChatGPT depends on making existing data knowledge accessible to the agent, since queries can rely on complex metrics and definitions scattered across tools and people. A semantic layer can capture key entities, metrics, filters, dimensions, open questions, and pitfalls, then be reviewed by data experts and packaged as a shared plugin. Teams can also bring in semantic layers they have already built, including ones in Snowflake Cortex or Databricks Genie. Doing this well requires focused, sustained investment from a team and organizational attention.

Key points

Complex queries and metric definitions can make it difficult to trust an agent’s data answers.
Relevant data knowledge may be scattered across tools and people.
A data plugin can turn existing knowledge into a shared skill for the agent.
A semantic layer can describe key entities, metrics, standard filters, and dimensions.
The semantic layer may also document open questions and pitfalls.
A data team familiar with the data should review and refine the layer before sharing it.
A well-contained domain is a good place to try this approach.
Existing semantic layers, including those in Snowflake Cortex or Databricks Genie, can be brought in rather than recreated.
Developing this well takes focused effort, continued investment, and organizational attention.

Chapters

Start	Chapter	Summary
00:03	Making data knowledge accessible	Complex queries depend on knowledge scattered across tools and people, so that knowledge must be accessible to the agent.
00:34	Building and sharing a semantic layer	A semantic layer captures entities, metrics, filters, dimensions, and caveats for review by the data team and sharing as a plugin.
01:04	Reusing existing layers and investing in quality	Teams can bring in semantic layers they already have, but making the approach work well takes sustained, focused investment.
01:35	Organizational attention and a teaser	The speakers emphasize the need for organizational attention and hint at a future product announcement.

Quotes

So it’s all about making it accessible to the agent. — 00:03

It has some information on like open questions, pitfalls. — 00:34

It can just learn from what you already have. — 01:04

But to do this really well it does take focused effort — 01:04

but you do need that attention. — 01:35

Concepts and entities

ChatGPT Work, data plugin, semantic layer, skill, Snowflake Cortex, Databricks Genie

Full transcript (original language)

00:03 I’m sure many of you in the audience are probably wondering how we would even trust chat GPT to find and query the correct data. Then even simple data questions can have very complex queries or metric definitions underpinning them and that knowledge can be scattered across tools, across people and if it’s not accessible to the agent then it’s not going to be able to make the right decision. So it’s all about making it accessible to the agent. And with the data plugin we can take that existing knowledge and then turn it into a shared

00:34 skill. So this this site semantic layer it contains information about key entities, key metrics, different standard filters or dimensions. It has some information on like open questions, pitfalls. Someone from your data team who is or who is familiar with this data should like review and refine this and then when they think it looks good they can package this up into a plugin and share that across your company. >> So if you have a relatively well contained domain this is a great skill

01:04 to try out. The second is that if your team has already invested in creating a semantic layer whether that’s as a skill or it might exist in Snowflake Cortex or Databricks Genie you can bring that in so you don’t need to create something that new. It can just learn from what you already have. >> But to do this really well it does take focused effort and like a team of people who is focused on this and continuing to invest in it. >> Yeah you have to really think this is important for the organization. I think

01:35 we can attest that it has been but you do need that attention. We will have another another announcement around a product in this space. I’ve said too much but anyway stay tuned.