The martech reset: the stack is collapsing into the data layer

Three eras in ten years

The customer data platform has been rebuilt twice. The packaged era, roughly 2016 to 2020, asked you to copy everything into a proprietary store, resolve identity there and sync audiences out. The composable era that followed kept the warehouse as the source of truth and layered activation on top, so the data stopped being duplicated. The era now opening is agentic. Agents resolve identity, assemble audiences and run always-on programmes directly against the data wherever it lives, and the location of the data matters less than who is allowed to decide what happens next.

Gartner frames the choice in front of CMOs as platformization or agentification. Platformization means the customer data platform becomes the floor of a large application suite. Agentification means it stays deliberately thin and hands execution to autonomous agents working on a shared data layer. The two roads look nothing alike on a slide. They agree on one thing that matters more than their differences. The data layer is the asset. The application layer is allowed to flex.

Automotive already lives this problem

Most dealer groups do not have a martech stack so much as a sediment. A CRM from one era, a DMS that predates modern marketing entirely, a website vendor, a chat tool, an equity mining product, a direct mail house, a service marketing vendor, an agency running paid, and a reputation platform. Each holds a partial copy of the same customer, each defines a sold vehicle slightly differently, and each was sold on the promise that it would finally unify everything.

The tools are not the expensive part. The reconciliation is. Every integration is a place where a customer can be counted twice, suppressed in one system and mailed by another, or offered a lease on a vehicle they bought in July. Add an autonomous agent to that and the reconciliation gap stops being a reporting inconvenience and becomes customer-facing.

The tools are not the expensive part. The reconciliation is.

What to hold, what to rent

The practical version of the strategy is a sorting exercise. Some capabilities are worth owning outright, because switching them is painful and because they determine the quality of everything downstream. Others should be treated as replaceable, because the market will keep producing cheaper and better versions of them.

• Own the identity layer. Which records are the same person, the same household, the same vehicle. This is the part nobody can rebuild for you quickly.
• Own consent, suppression and frequency. These are business rules, not vendor features, and they need to apply across every channel at once.
• Own the outcome record. What actually happened in the store, written back and linked to the activity that preceded it.
• Rent the creative and execution surfaces. Channels, templates, bidding and copy generation are converging on commodity fast.

The test for any renewal is no longer whether the tool does the job. It is whether leaving that tool would take your customer data with it. If the answer is yes, you are not buying software. You are handing over the asset.

Questions worth asking a vendor in 2027

Ask where the data physically lives, and whether you can read it without the vendor. Ask how identity is resolved, and whether the logic is inspectable or a black box. Ask what happens to your suppression rules when a new channel is added. Ask for provenance on every enrichment, because an agent acting on an attribute nobody can source is a compliance conversation waiting to happen. Ask what the exit looks like, in writing, before the relationship is good.

None of these are AI questions. That is the point. The market is consolidating around governed data foundations rather than around cleverer applications, and the organizations that come out of this reset in a strong position will be the ones whose stack got simpler in the places that matter and stayed flexible everywhere else.

Written by Mercy. Automotive Marketing Strategy, ClicMotion.