Integration · Operations and Data

Bring Databricks features and predictions into Agentforce

Databricks is a lakehouse: Delta tables holding both raw and refined data, Unity Catalog governing who can touch it, and the ML features and model outputs your data science team produces on top. When Emerge Digital connects it to Salesforce Agentforce, an agent can read not just the data but the predictions — a churn score, a propensity, a recommended segment — that already live there. Unity Catalog still decides access, so the agent works from the same governed lakehouse and the same model outputs your analysts and pipelines rely on.

All integrations

What this unlocks

Agents read model outputs computed in Databricks — a churn likelihood, a lead score, a next-best-offer — so a recommendation carries the reasoning of a model, not just a lookup.
Delta tables in the lakehouse are available to agents through Unity Catalog, so refined and raw data can be reasoned over under one governance model.
Feature values your team engineered for models become inputs an agent can cite, keeping the agent's view consistent with what the models were trained and scored on.
Unity Catalog permissions scope what each agent may read, so a prediction stays visible only to the agents and people entitled to see it.

In the customer journey

Act on a churn score, not a hunch

A success agent looking at an account reads the churn prediction scored in Databricks and adjusts its outreach, so the intervention lines up with what the model flagged rather than a manual guess about risk.

Segments that match the model

When an agent suggests who to target, it reads the propensity features and segment assignments from the lakehouse, so the recommendation reflects the same feature set the data team built rather than an ad-hoc rule.

Databricks earns its place across Engage, Convert, and Optimize, because that is where model outputs change what an agent should do. A propensity score can shape how an agent engages a prospect; a churn prediction can steer a retention play; and as pipelines rescore, the agent's next recommendation reflects the latest model run rather than a fixed rule set.

How Emerge integrates Databricks

Emerge Digital integrates Databricks with Agentforce as a consulting engagement — never a self-install app. We identify which Delta tables, features, and model-output tables your agents need, configure Unity Catalog access so each agent reads only what it should, and build the retrieval and actions that let an agent use predictions in a conversation. Where a model output drives a high-stakes step, we keep a person in the loop to confirm the call. Governance stays with Unity Catalog, not with prompt wording; you own the lakehouse, and we make its data and predictions usable inside your agents.

How we structure an engagement

Related integrations

FAQ

Can agents use our model predictions, or only raw data?

Both. A common reason to connect Databricks is exactly the model outputs — scores and predictions your team already computes — so agents can act on them. Emerge maps the feature and output tables you want exposed.

Is this a pre-built Databricks connector we self-install?

No. Connecting Databricks to Agentforce is a services engagement. Emerge configures Unity Catalog access and builds the retrieval and actions around your lakehouse rather than shipping an app you install.

How is access to lakehouse data controlled?

Through Unity Catalog, not the prompt. An agent can only read the tables, features, and outputs its Unity Catalog permissions allow, and we keep a person in the loop where a prediction drives a consequential decision.

Ground your agents in Databricks.

Tell us what your agents need to read and write in Databricks, and we'll design the integration and the governance around it.

Talk to the practice

Prefer email? Write to the practice instead.