A managed and governed approach to context is vital to the success of Analytics Agents in the Enterprise

These days there's a lot of talk about context, and suddenly every AI company is a "context company". At WisdomAI, however, we've been focused on building the Adaptive Context Engine for three years, before context was "cool." We don't just talk about the importance of context. We give data teams the practical tools to build, maintain, and scale the enterprise context that makes Analytics Agents work.

This week, I'm excited to bring you a two-part roundup of our Spring releases. In this article, we'll focus on new features and enhancements that help admins accelerate WisdomAI domain setup, govern context, connect to more data sources, and optimize token cost. In our next edition we'll focus on new features that help WisdomAI end-users get faster, better insights.

Build context from artifacts

One of the most time-consuming parts of setting up a new analytics domain is manually defining table descriptions, joins, metrics, and business rules. But most of that knowledge already exists somewhere – in your dbt YAML files, LookMLs, internal documentation, or historical query logs. Getting it into WisdomAI has usually meant entering it by hand.

Our new Context Extractor changes that. Simply upload your existing documentation: dbt YAML files, LookML models, PDFs, plain text docs, or SQL query logs. Then, WisdomAI automatically analyzes the documentation to propose structured context: Table and column descriptions, metric definitions with SQL, join relationships, business rules, and example text-to-SQL pairs for model training.

Every extracted item appears as a reviewable suggestion in your Domain's Knowledge tab, showing the exact source content that generated it. You can accept, reject, or edit each suggestion before it goes live. For suggestions that involve SQL, WisdomAI validates the logic by running a test execution against your warehouse before surfacing it.

For teams migrating from Looker or Tableau, this is particularly powerful. Existing data modeling logic doesn't have to be rebuilt from scratch. For teams starting fresh, it dramatically compresses the time between connecting a data source and producing accurate, trusted answers and reliable insights.

Tabular extraction from unstructured data

A lot of business-critical data lives outside of WisdomAI. such as contracts, invoices, research reports, compliance filings, and procurement documents. Analysts have historically had two choices for unstructured data analysis: manually enter the data into a spreadsheet or ignore it entirely when building analysis.

WisdomAI can now extract structured data from unstructured document collections and turn it into a queryable table. Upload a folder of PDFs, Word docs, PowerPoints, or images. Define the fields you want using plain-English prompts. Preview extraction on a sample before committing, and then run the full job. The extracted unstructured data table sits alongside your warehouse or third-party app data, allowing you to unlock insights via joins between the queryable table and transactional data tables in a single query.

For example, procurement teams can now analyze contract terms originally stored in an S3 bucket alongside structured transactional data from the financial system for a new class of insights. Or, Finance can process invoice line items without manual entry, and compliance teams can analyze thousands of filings in just minutes.

Metrics and Derived Columns

Getting metrics right is one of the most time-consuming parts of building a reliable analytics domain. You know what KPIs like "profit margin" means to you, but translating that into correct SQL, pointed at the right source table, and validating against real data is a challenge.

WisdomAI's new Metrics and Derived Columns feature lets you define business logic in natural language. Describe what you want, such as "calculate profit margin as revenue minus cost divided by revenue," and WisdomAI generates the SQL, identifies the correct source table, and tests the expression against your data to validate before saving. You can review and edit the generated SQL before it goes live.

WisdomAI automatically distinguishes between metrics (aggregated values across rows, such as total revenue) and derived columns (row-level calculations, like combining first and last name) based on how you describe them. Domain administrators and business analysts can now define accurate business logic without needing to be SQL experts.

Real-time internet search as a data source

WisdomAI now supports Internet search as a configurable data source on any domain. When enabled, WisdomAI automatically supplements analysis with real-time information from the web, including industry benchmarks, market trends, competitor data, and any publicly available material — even historical weather patterns. Users can explicitly ask WisdomAI to do an Internet search in a prompt, but by default the system decides when to invoke web search based on the question, giving depth and nuance to every answer.

Admins control exactly which websites WisdomAI is allowed to search. They can open a domain to the full Internet, restrict to an approved list of domains, or block specific domains. Results are returned with inline citations and a full sources list so users can verify where the information came from.

Teradata connector

Teradata Vantage powers some of the most mission-critical data environments in very large enterprises. Until now, organizations running Teradata had no direct path to natural-language analytics without replicating their data to a separate warehouse. WisdomAI now connects natively to Teradata Vantage, both on-premises and cloud deployments.

The connector handles automatic schema discovery, generates Teradata-dialect SQL, including correct window function handling, and executes all queries in read-only mode. Enterprise teams on Teradata can now enable self-serve analytics across their existing Teradata sources without data pipelines, migration, or middleware.

Google Cloud Spanner support

Google Cloud Spanner is increasingly the data layer of choice for GCP-native engineering teams building globally distributed applications. The challenge is that bringing analytics to Spanner typically requires ETL to a separate warehouse.

WisdomAI now connects directly to Google Cloud Spanner. Whether your Spanner database uses Google Standard SQL or the PostgreSQL interface, WisdomAI automatically detects the dialect at connection time and handles it without any configuration from admins. Schema crawling, natural-language queries, and read-only execution all work out of the box.

Smarter data management: partition and view awareness

Running a query against a partitioned table without a partition filter is one of the most common and expensive mistakes in cloud data warehouses. In Google BigQuery, for example, queries can trigger a full-table scan that can fail entirely on partition-required tables, generating wasteful compute charges.

WisdomAI now automatically detects partition columns during schema crawling for BigQuery, ClickHouse, Databricks, and Redshift. Generated SQL includes appropriate partition filters by default, preventing both errors and unnecessary scans. Background crawls also run more efficiently, scanning only recent partitions of large tables rather than scanning everything on every run. WisdomAI distinguishes between base tables, views, and materialized views and adjusts query planning accordingly with no additional configuration required.

Smarter data crawling

Scheduled crawls that scan every text column in a large schema generate a lot of warehouse queries with very little return. Most columns don't contain values worth indexing for filter drop downs.

WisdomAI now uses AI to identify which columns contain meaningful categorical values, like customer names, product names, and status fields, and crawls only those. When you add a new filter to a dashboard or enable lookup on a column, WisdomAI immediately triggers a targeted crawl for exactly that column, with values appearing within 3 to 4 minutes rather than waiting for the next scheduled run.

Teams with large schemas will see meaningful reductions in background crawl volume and associated warehouse costs. You can still manually enable value crawling on any column from domain settings, and that setting takes priority over automatic detection.

Federated Agentic Intelligence: WisdomAI is agnostic at the insights layer

Taken individually, each of these updates solves a specific friction point around building enterprise context or connecting to more data sources. Our product strategy is to be data source agnostic at the insights layer, and to provide data teams in the enterprise with the tools to automate, manage, and govern enterprise context as the fuel that makes Agentic Analytics successful.

With WisdomAI, you get Conversational BI, AI-powered dashboards, and Analytics Agents providing the highest quality insights: the most accurate, most cost-efficient, and most deeply relevant agentic intelligence across all your data.

See what WisdomAI can do for you

If you want to see how these capabilities work against your actual data sources, including Teradata, Spanner, unstructured document collections, or Snowflake, Databricks, and more, we are happy to work with you. Request a demo and let us put WisdomAI to work on your highest-impact analytics use cases.