Some tasks are well suited to automation, and Agentic AI is improving all the time when it comes to various use cases, not least code generation.
But certain roles are more important than ever in relation to Data Solution delivery, and the one that keeps coming to the top of my list is Data Modeling.
Whether it’s Conceptual Models, Ontologies, or Semantic Layers, there is no role better placed than the Data Modeling role to help get things right.
How does Agentic AI get the elusive “context” that is a fundamental ingredient for trustworthy AI?
How does it get to the tacit knowledge that people may not even realize is tacit knowledge?
The synonyms in the business vernacular?
The relationships between database tables that, if you were depending on inconsistent naming standards alone to decipher, may likely remain undeciphered!
Take semantic layers on the Snowflake Data Platform as an example. Semantic Views are configured using e.g.:
– Dimensions
– Facts
– Metrics
– Relationships
– Verified Queries
– Synonyms
Semantic Layer configured in Cortex Analyst on the Snowflake Platform. Credit: Dan Galavan
Relatively speaking, the technical configuration is the easy part. In fact if you use the Snowflake Cortex Code Agent (a.k.a. CoCo) to automatically configure a semantic view using YAML based on the SNOWFLAKE_SAMPLE_DATA database, it will do a very good job.
But this is based on a well defined, well named, and sanitized database. And that’s not unusual for sample databases, it’s actually the norm.
Out in the ‘real world’ however, things are less clear, less defined, less deterministic. More stochastic.
And that’s it in a nutshell. To summarize, I’ve come up with a quote (yes, a human generated quote!):
“In all the AI excitement / hype / hyper pace of change, Data Models are an anchor to true data value”!
But certain roles are more important than ever in relation to Data Solution delivery, and the one that keeps coming to the top of my list is Data Modeling.
Whether it’s Conceptual Models, Ontologies, or Semantic Layers, there is no role better placed than the Data Modeling role to help get things right.
How does Agentic AI get the elusive “context” that is a fundamental ingredient for trustworthy AI?
How does it get to the tacit knowledge that people may not even realize is tacit knowledge?
The synonyms in the business vernacular?
The relationships between database tables that, if you were depending on inconsistent naming standards alone to decipher, may likely remain undeciphered!
Take semantic layers on the Snowflake Data Platform as an example. Semantic Views are configured using e.g.:
– Dimensions
– Facts
– Metrics
– Relationships
– Verified Queries
– Synonyms
Relatively speaking, the technical configuration is the easy part. In fact if you use the Snowflake Cortex Code Agent (a.k.a. CoCo) to automatically configure a semantic view using YAML based on the SNOWFLAKE_SAMPLE_DATA database, it will do a very good job.
But this is based on a well defined, well named, and sanitized database. And that’s not unusual for sample databases, it’s actually the norm.
Out in the ‘real world’ however, things are less clear, less defined, less deterministic. More stochastic.
And that’s it in a nutshell. To summarize, I’ve come up with a quote (yes, a human generated quote!):