Databricks is right: AI has a context problem. In automotive, context is identity.

At Advertising Week New York, two of Databricks’ media and advertising leads made a case every data team should hear: the next wave of AI value comes from bringing AI to your data, not your data to AI.

On the Great Minds podcast they repeated a line from Databricks CEO Ali Ghodsi that sums it up:

“LLMs don’t have an intelligence problem. They have a context problem.”

We agree, and we would go one step further. In automotive, context is identity and intent. A model can only reason about the customers it can tell apart.

Context starts with knowing who’s who

If one buyer appears as four records across the CRM, the DMS, the service lane and a web form, every model downstream reasons about four strangers. Loyalty looks weak, attribution breaks and the best customer in the store looks like four mediocre leads. No amount of model intelligence fixes a foundation that splits one person into four, or merges a father and son into one impossible customer.

Control has to extend to agents

The episode framed governance as something that now covers agents, not just people. That is the right frame. Agents that prepare audiences and orders need the same guardrails a careful analyst would follow: re-identification only with the data owner’s authorization, regulated data used only as its license allows, and a person approving what leaves the building.

Prove the next dollar

The industry is moving from reporting what happened to predicting what the next dollar will return. Automotive has an advantage here that most categories don’t: a hard outcome. A sale lands in the DMS, tied to a vehicle. When identity is resolved, the loop from audience to campaign to sold vehicle can close, and the next dollar can be argued with evidence instead of adjectives.

Bring the AI to the data

Customers told Databricks to run marketing where they run the business rather than copying golden profiles into yet another system. DarkMath was built on the same instinct. Our software can run inside a client’s environment, connect data on anonymous IDs and keep re-identification behind the client’s gate.

Where horizontal platforms stop

Horizontal platforms supply the plumbing, and they do it well. The vertical context is the part each industry has to bring: the household’s vehicles, the market around the store, the signals that someone is in market, and the regulated data rules that govern how all of it may be used. That is the layer DarkMath provides.

Databricks brings AI to your data. DarkMath brings meaning to your identity.

Sources: What Will the Next Advertising Dollar Deliver? (Advertising Week, Great Minds) and Databricks CustomerLake announcement.