𤯠Treasure Dataās 2026 Agentic AI: The CDP Upgrade Marketers Didnāt See Coming
- Martexh Expert

- Mar 25
- 2 min read
I recently had the chance to test Treasure Dataās new 2026 Agentic AI capabilities in a sandboxāand honestly, I wasnāt expecting to be this impressed.
At first glance, the AI agent feels like a standard chatbot. But once you start using it, you realize itās much more than that.
I was able to have a full back-and-forth conversation about audience segmentationāeverything from which attributes to include to how to structure entire customer journeys. Even better, with a single click, I could turn those ideas into actual segments inside the platform.
All I had to do was describe what I wanted. For example:
ZIP code targeting
Site visits within the last month
More complex attributes like customer type or gender
The AI didnāt just understandāit helped refine and build.
Where things really got interesting was with custom attributes. I asked how to create a calculated metric like:
Sales Ć products purchased Ć· visits over the last 6 months
Not only did it generate the SQL query, but it also walked me through exactly how to implement it in the platform.
Now, itās not perfectāand thatās probably a good thing.
When I asked it to execute the SQL directly, it couldnāt. And honestly, that limitation makes sense. Giving AI full control to modify platform code could create more risk than reward.
That said, after a bit of back-and-forth (it took about three tries), the AI delivered a clean, fully usable SQL queryāno missing pieces, no unnecessary comments, no hallucinations. Just something I could copy, paste, and run.
Thatās the real value here.
This isnāt just AI for the sake of AIāitās a practical tool that simplifies one of the most complex parts of working in a CDP: configuring data, segments, and calculations.
Big credit to Treasure DataĀ for pushing the boundaries and making sophisticated data work more accessible for marketers.
If this is where Agentic AI is headed, CDPs are about to get a whole lot more user-friendly.




Comments