Legacy SQL to AI-Ready Relationship Evidence

You have a SQL Server database from 2012. It has hundreds of tables, inconsistent naming conventions, and documentation that was last updated in another era. Your CEO wants AI. Your board wants AI. Your competitors have AI. But your data is not ready. Or is it?

Phase 1: Discovery

KynticAI connects with approved read-only credentials or an authorised one-off import. It scans schemas, maps column types, identifies likely relationships, and generates a metadata profile. No ETL project has to begin before the buyer can see the shape of the relationship layer.

The output is a semantic map: a visual representation of the data estate showing which tables contain customer context, transactional history, and operational events. Fields are tagged with semantic hints such as "this looks like an email address", "this looks like a product interest", or "this looks like an account registration event".

Phase 2: Selector Configuration

Selectors define how raw database fields translate into relationship evidence. The mapping engine can handle direct field mappings, controlled vocabulary mappings, threshold classifications, weighted scoring, and custom formula metrics.

DirectFieldMapping: The "email" column becomes a contact email item.
StringToEnumMapping: Free-text values like "Hot/Warm/Cold" become an engagement score.
ThresholdClassification: Numeric values are classified into bands such as low, medium, and high value.
WeightedScoring: Multiple fields combine into a composite relationship score.
FormulaMetric: Custom calculations turn source data into derived relationship signals.

Phase 3: Relationship Assembly

With selectors configured, the relationship layer begins creating facts. Each fact includes the item, the resolved relationship, a confidence score, and provenance that traces back to the source table, column, event, or import.

A customer item might include churn risk, conversion probability, preferred channel, last web search, account registration path, and the product they inspected before making contact. The point is not to flatten the data into one giant table. The point is to preserve the path.

Phase 4: Integration and Validation

The Rust relationship engine compares the current path with millions of similar paths and creates top-example JSON. Your approved LLM then explains the next task: send a follow-up email, ask the user to register, offer a product guide, route to a human, or hold back because the evidence is weak.

As business outcomes occur, the flywheel attributes the result back to the relationship evidence that influenced it. Over time, the engine learns which legacy SQL signals actually matter for the customer's own goals.

What You Did Not Have to Do

You did not build a new data warehouse. You did not hire an ETL engineering team before the first useful question could be asked. You did not migrate to a new database. You did not rewrite your applications. You connected KynticAI to what you already had, and the old SQL estate started producing relationship evidence an AI system could actually use.