Connect AI to
Your Data. Securely.

The classic sales intelligence query. Account managers pull this to prep for client calls. Sales leadership uses it for pipeline reviews. Works against any CRM or ERP database connected through DreamFactory.

The executive's first 'aha' moment with AI. A CFO or CEO asks a plain-English question and gets an answer in seconds instead of requesting a report that takes days. This is typically the demo that sells AI projects internally.

Quarter-end reporting in seconds instead of hours. Finance teams use this to validate revenue numbers. Executives use it for board prep. The GROUP BY runs server-side, so AI gets summary data, not millions of raw records.

Quality teams in manufacturing use this to catch production anomalies. Healthcare uses it for patient safety outlier detection. The two-step pattern (aggregate for stats, then query for outliers) is a core QA workflow.

Production-grade query execution. When the query is too important to leave to AI's judgment, a stored procedure ensures the exact same logic runs every time. This is the backbone of the employee evaluation pipeline and every finance automation.

The write-back that makes AI pipelines real. AI doesn't just read data; it creates records. Sales pipelines write leads back to CRM. Finance pipelines write match records to ERP. HR pipelines write evaluation drafts to HRIS.

After acquisitions, data lives in multiple systems with no single source of truth. This pattern queries both, compares results, and finds discrepancies. Essential for data migration validation and ongoing reconciliation.

Legacy databases and NoSQL systems often lack proper foreign key relationships. Virtual FKs let AI follow data relationships that exist logically but not physically in the schema. Critical for legacy system AI enablement.

Before any pipeline write-back step, AI needs to know the exact column types and constraints to format data correctly. This prevents write failures from type mismatches or null violations.

When AI needs to understand the full query syntax available for a service. Particularly useful for technology teams building custom integrations or documenting their AI pipeline architecture.

Operations teams scanning for new data files on SFTP servers. Distribution companies checking for incoming EDI files. Manufacturing teams looking for PLC data exports. The first step in any file-based data pipeline.

AI reads and summarizes files without anyone downloading, opening, or reformatting them. Finance uses it for bank statement processing. Operations for production reports. The auto-detection handles CSV, JSON, and text natively.

The classic integration pattern: read a file, transform the data, write it to a database. This replaces manual data entry and custom ETL scripts. Distribution companies use it for EDI processing. Manufacturing for supplier data imports.

The output step of many AI pipelines. Finance writes exception reports to S3. HR writes evaluation summaries to shared drives. Sales writes lead lists for CRM import. Every department has a 'generate and save' workflow.

Compliance teams comparing this month's regulatory report against last month's. Financial services auditing quarterly filings. Healthcare reviewing patient outcome trends across reporting periods.

DevOps and IT teams managing application configuration across environments. Read a config, modify a setting, write it back. Particularly useful for managing DreamFactory service configurations programmatically.

Processing dozens of incoming files in one pass. Distribution companies handling daily EDI batch files. Manufacturing processing shift reports from multiple facilities. AI iterates through a directory and extracts insights from each file.

The output half of every reporting pipeline. Query the database, format results, save to cloud storage. Finance exports reconciliation reports. Sales exports CRM extracts. Operations exports production summaries.

After processing, old files need cleanup. Role-based permissions ensure only authorized users can delete files. Audit logging tracks every deletion. Essential for regulated industries with data retention policies.

Manufacturing quality teams analyzing images of product defects. Healthcare reviewing medical imaging metadata. AI receives images and audio as base64 for analysis without downloading files to local machines.

The first security step for any AI deployment. Sales gets read access to CRM tables. Finance gets access to AR data. The shop floor gets production data only. Each role sees exactly what it should and nothing more.

The enterprise security requirement: every AI query runs under the real user's database credentials, not a shared service account. This preserves existing database-level audit trails and permission models.

Territory-based data isolation for sales teams. Department-scoped access for HR. Region-restricted views for operations. Server-side WHERE clauses that AI cannot bypass or override, enforced on every request.

Setting up the secure connection between AI tools and DreamFactory. OAuth 2.0 + PKCE ensures tokens expire and refresh automatically. Critical for ChatGPT and Claude Desktop integrations.

Each AI tool gets its own API key with its own permissions. Claude gets one key, ChatGPT gets another, Cursor gets a third. If one key is compromised, only that tool's access is affected.

HIPAA and SOC 2 compliance requirement: mask SSN, salary, patient IDs, and other sensitive fields so AI can query employee or customer data without ever seeing PII. Essential for healthcare and financial services.

Enterprise SSO integration: your existing Azure AD groups automatically map to DreamFactory roles. When a new employee joins the Sales group, they immediately get the right AI data access without manual provisioning.

The write-safety pattern: AI can EXECUTE stored procedures but cannot directly INSERT, UPDATE, or DELETE from tables. All write operations go through pre-approved, tested code paths. This is how production AI pipelines stay safe.

For organizations using Okta, Auth0, or Azure AD as their identity provider. Tokens are validated server-side using the provider's public keys. No shared secrets to manage or rotate.

Token economics and cost control: limit how many requests each AI tool can make per minute. Prevents runaway AI queries from overwhelming your database. Different limits for different roles (admin vs. standard user).

Data teams exploring their Snowflake warehouse structure before building analytics pipelines. See every schema, table, and column type. Particularly useful when onboarding new analysts or auditing data warehouse coverage.

Business analysts querying Snowflake through natural language instead of writing SQL. The 500-record limit and server-side filtering ensure AI gets precisely the data it needs without pulling entire tables.

Revenue reporting, customer analytics, and operational metrics computed server-side in Snowflake. No raw data transferred. AI receives summary numbers and generates insights from aggregated results.

Production-grade Snowflake queries wrapped in stored procedures for repeatability and audit compliance. Finance teams use these for month-end reporting. Data teams use them for ETL validation.

The common enterprise pattern: analytics live in Snowflake but operational data lives in SQL Server or Oracle. AI queries both through DreamFactory and joins the results, creating a unified view without data migration.

Complex analytical workflows: first aggregate to identify top customers, then drill into their order patterns. AI chains multiple queries to build layered insights, mimicking the analytical process a human analyst would follow.

Inventorying available Snowflake procedures before building AI pipelines. Particularly useful for new team members learning what pre-built analytics and reporting procedures are available.

AI discovers the data model first, then builds reports intelligently. This two-step pattern ensures AI queries are schema-aware, reducing errors and producing more accurate reports.

DreamFactory adds an access control layer on top of Snowflake's native permissions. Analytics teams see aggregated data. Executives see everything. Individual contributors see only their department's data.

Writing AI-generated results back to Snowflake: processed scores, aggregated metrics, or AI classifications. Completes the pipeline loop where AI reads from Snowflake, processes data, and writes results back.

The production pattern from the employee evaluation use case: stored procedure collects data, DreamFactory serves it, Python script sends to LLM, returns AI summary. This is how a DreamFactory customer processes 8,000 evaluations in 16 seconds.

Catch bad data before it reaches your database. Validate required fields, check format consistency, enforce business rules. Every write-back step in an AI pipeline should have input validation as a safety net.

The alert mechanism for AI pipelines. When a new high-priority AR record is created, Slack the collections team. When a defect pattern is detected, alert the quality manager. Event-driven notifications without polling.

Regulatory compliance requirement for healthcare (HIPAA), financial services (SOC 2), and government (FedRAMP). Every data change captured with old values, new values, user identity, and timestamp.

Reshape data for specific consumers. Format Oracle date fields for a React frontend. Convert ERP codes to human-readable labels for AI consumption. Transform XML responses to JSON for modern applications.

The multi-source pipeline pattern: pull internal data, call an external API for enrichment (credit scores, weather data, market signals), merge the results, and save. This powers lead scoring and risk assessment workflows.

Build your own AI pipeline as a single API endpoint. Database query, prompt construction, LLM call, and response formatting all wrapped in one scripted service. This is the pattern behind the executive briefing generator.

Keep databases in sync across systems after mergers or integrations. The urllib timeout pattern prevents PHP-FPM worker pool deadlock, which is critical for production reliability.

The output step for executive dashboards, compliance reports, and operational summaries. Query data, format as PDF/CSV/JSON, upload to cloud storage. Automate the 'generate and distribute' workflow.

The compliance safety net: even if RBAC allows access to a table, this script strips sensitive fields before they reach AI. Critical for healthcare patient data, financial PII, and government personnel records.

Wrap any external API with DreamFactory security: RBAC, rate limiting, and audit logging. Your team uses one security model for both internal databases and external APIs.

Connect to SaaS platforms (Salesforce, HubSpot, Stripe) through DreamFactory, which handles token refresh automatically. No more expired token errors in production pipelines.

Secure internal API access through DreamFactory roles. Engineering gets full access. Sales gets read-only. Partners get a subset. One gateway, multiple audience-specific access patterns.

Add correlation IDs, auth tokens, or tracing headers to every proxied request automatically. Essential for distributed systems observability and debugging production issues.

The enrichment pattern: combine external API data (weather, market signals, credit scores) with internal database data in one response. This powers the lead targeting and risk scoring use cases.

Route AI model calls through DreamFactory for cost control and token tracking. Rate limit per department. Track token usage per role. The model routing pattern described in the executive dashboard use case.

Unify access to multiple external APIs under one security layer. Instead of managing credentials for each API separately, DreamFactory handles all authentication and access control centrally.

Auto-generate documentation for your proxied APIs so AI agents can discover endpoints and parameters. The more AI knows about available APIs, the smarter its query planning becomes.

Debug API integrations by logging every request and response. Monitor external API reliability. Track usage patterns. Essential during pipeline development and ongoing production monitoring.

Production monitoring for API health. Track response times, error rates, and throughput across all proxied services. Detect degradation before it impacts AI pipeline performance.