Snowflake Cortex AI
Snowflake Cortex AI is Snowflake's set of LLM functions, search, natural-language-to-SQL, and agent services that run inside the Snowflake data cloud next to a company's governed data.
Definition
Cortex AI lets teams call large language models from SQL through AI functions such as AI_COMPLETE, AI_CLASSIFY, AI_EXTRACT, and AI_PARSE_DOCUMENT, with models from Anthropic, OpenAI, Meta, Mistral, Google, and Snowflake's own Arctic family depending on region and input type. Cortex Analyst turns natural-language questions into SQL against semantic views that define business metrics and relationships, while Cortex Search provides retrieval over unstructured text for RAG and enterprise search. Cortex Agents orchestrate multi-step work across those tools plus custom functions and stored procedures, and they power Snowflake Intelligence, a conversational app for business users. Snowflake has been adding MCP connectors, agent skills, a sandboxed code execution tool, and agent evaluations, with some of these still in preview. Access to each model is governed by Snowflake roles. For construction companies, Cortex AI usually comes into play once ERP, project management, and field data already land in Snowflake.
In Depth
Snowflake Cortex AI is a set of AI services inside a data platform that many large contractors and owners already use for reporting. Its appeal is proximity: the models run next to the data, under the same access controls, so a team does not need to copy job cost or payroll records into a separate AI tool. At the simplest level, an analyst can call AI_COMPLETE or AI_EXTRACT inside a SQL query to summarize text fields or pull values out of documents staged in Snowflake.
The higher-level services are where construction data teams spend their time. Cortex Analyst answers plain-language questions by writing SQL against semantic views, which are definitions of metrics like committed cost, earned hours, or percent complete that the data team maintains. Cortex Search indexes unstructured text, such as daily logs or meeting minutes, for retrieval. Cortex Agents combine the two with custom tools and power Snowflake Intelligence, the chat interface business users see. The quality of answers depends heavily on the semantic views: if cost codes, phases, and project identifiers are inconsistent across ERP and project management sources, the agent inherits that inconsistency.
For a contractor weighing Cortex AI, the real prerequisite is a working warehouse with construction data already modeled, which is why it often appears alongside connector and modeling vendors. Evaluate credit consumption on realistic workloads, check which models are available in your Snowflake region, and decide who owns the semantic views, because that ownership determines whether field and finance teams trust the answers.
Examples
Using AI_EXTRACT to pull invoice numbers, amounts, and retainage from subcontractor pay application PDFs staged in Snowflake
Setting up Cortex Analyst on a semantic view of job cost data so an operations VP can ask which projects are over budget on labor
Running AI_COMPLETE over a year of daily log text to summarize recurring weather and inspection delays by project
Frequently Asked Questions
Cortex AI lets teams call large language models from SQL through AI functions such as AI_COMPLETE, AI_CLASSIFY, AI_EXTRACT, and AI_PARSE_DOCUMENT, with models from Anthropic, OpenAI, Meta, Mistral, Google, and Snowflake's own Arctic family depending on region and input type. Cortex Analyst turns natural-language questions into SQL against semantic views that define business metrics and relationships, while Cortex Search provides retrieval over unstructured text for RAG and enterprise search. Cortex Agents orchestrate multi-step work across those tools plus custom functions and stored procedures, and they power Snowflake Intelligence, a conversational app for business users. Snowflake has been adding MCP connectors, agent skills, a sandboxed code execution tool, and agent evaluations, with some of these still in preview. Access to each model is governed by Snowflake roles. For construction companies, Cortex AI usually comes into play once ERP, project management, and field data already land in Snowflake.
Using AI_EXTRACT to pull invoice numbers, amounts, and retainage from subcontractor pay application PDFs staged in Snowflake. Setting up Cortex Analyst on a semantic view of job cost data so an operations VP can ask which projects are over budget on labor. Running AI_COMPLETE over a year of daily log text to summarize recurring weather and inspection delays by project.
More AEC Tools Terms
View allAdd AI to your Snowflake Cortex AI workflow
Nomic is purpose-built AI for architecture, engineering, and construction. Connect your project data and start getting answers in minutes.