We now open the DSE Data Intelligence Agent, moving from exploration to intelligence-assisted modernization.

The provider is local Ollama.

The selected model is Gemma 3 4B, keeping the default AI path local and privacy-focused.

The Advisor is metadata-first and read-only.

We analyze the discovered sources.

DSE completes deterministic analysis independently of model enhancement, so useful findings remain available even if AI enhancement times out.

The Advisor summarizes discovered sources, objects, and fields.

It identifies structural repetition and potential overlap.

High-confidence matches remain separate from lower-confidence review candidates.

The Advisor recommends a canonical model and supported destination technologies.

We select a recommended destination using the Use action.

DSE can now create or refresh draft output targets.

The selected destination can also be staged for Generate Deployables.

AI does not write directly to an external system; the user reviews the route before execution.

We open the Output Target Composer.

Here, source fields become an explicit destination contract: objects, mappings, destination names, and types can be reviewed and adjusted.

DSE protects against duplicate mappings and preserves a reusable, reviewable target definition for the next deployment step.