Treasure Data became Treasure AI on April 20, 2026, after close to fifteen years — and the rename tracked a product decision, not a marketing one. The CDP stayed underneath; a decisioning layer and an activation layer went on top, with Treasure AI Studio orchestrating end to end. This four-part series brings in the product leaders behind that change, one vantage point at a time: the vision, the roadmap, the operational reality, and the design.
Check out the Treasure AI Studio sandbox for yourself
Part one, with Chief Product and Growth Officer Rafael Flores, ran on the main podcast feed. The other three are video conversations available here and on YouTube.
Featured: Rafael Flores, Chief Product and Growth Officer
Treasure AI Chief Product and Growth Officer Rafael Flores on evolving CDPs to activation platforms Published August 16, 2026 · The Agile Brand with Greg Kihlström®
Rafael “Rafa” Flores explains why a company that spent close to fifteen years building customer data infrastructure rebuilt itself as an agentic experience platform. The through-line is a problem he says went unsolved for over a decade: CDPs power the ESPs and marketing automation systems that generate revenue, but buyers could never attribute a dollar back to them. Adding orchestration and UTM-level tracking inside the platform is his direct answer.
What Flores argues in this conversation:
- The attribution gap, not the capability gap, was the CDP’s structural weakness. CDPs sit one layer back from activation, so the revenue got credited above them while the CDP carried the cost line.
- Fifteen years of data infrastructure is the moat, not the baggage. Warehouse-native entrants still have to solve real time — historically the crux for data warehouses — before they reach where CDP-native platforms already operate.
- Databricks entering the category validates the direction rather than threatening it. Customers already run Treasure AI alongside Databricks and Snowflake, so coexistence is a customer requirement.
- CDP deployment has collapsed from months to weeks. A use case can go live in roughly two weeks, with policy and governance — not data cleanup — as the pacing constraint.
- Slow data operations cost real campaigns. After the Silicon Valley Bank collapse, banks with CDPs reached customers the same day; one bank without that capability was still manually pulling and uploading sheets two days later and was blocked by regulators before its first send.
- AI resistance sits in the middle of the org, not the top. CMOs and C-suites push AI down while mid-management slows it, out of fear the push is a prelude to headcount cuts.
- “Show it to me now” is the buyer’s best vendor test. If a vendor needs time to assemble a demo, Flores says they don’t have the product.
About Rafael Flores — Chief Product and Growth Officer at Treasure AI. He has scaled SaaS product organizations at Meltwater (through IPO), Datanyze (through acquisition by ZoomInfo), ARM, and 6sense, and previously helped orchestrate Treasure Data’s $600M acquisition by ARM. A member of the Forbes Technology Council. Rafael Flores on LinkedIn
Treasure AI’s Kristen Zhou on how AI is changing how teams build products
Here’s a question to start with: “”agentic”” can mean a system that suggests something to you, or a system that does something and tells you afterward. Those are not the same product. They’re not even the same risk profile. And somewhere between them is where a roadmap actually gets decided.
This week we’re running a four-part series with the product team at Treasure AI — four product leaders , from four different seats. Today we’re in the seat where vision has to become a sequence: what gets built, in what order, and how much a system is allowed to do on its own. And all along the way, we’ll learn how AI is not only changing the type of products we build, but how product teams work entirely.
To help me get into that, I’d like to welcome Kristen Zhou, VP of Product at Treasure AI.
Treasure AI VP of Product Ops Michelle Morrison on AI’s impact on the product org
Here’s where I want to start: most of what breaks in a fast-moving product organization doesn’t break in the product. It breaks in the handoffs between the teams building it.
Speed is easy to measure. Whether everything you shipped quickly still adds up to one coherent product — that’s much harder to see, and it’s the thing that actually determines whether the speed was worth anything. That’s an operations problem before it’s a technology problem.
This week we’re running a four-part series with the product team at Treasure AI. Today we’re in the operations seat — the one where “”what’s next”” has to survive contact with how work actually gets done.
To help me dig into that, I’d like to welcome Michelle Morrison, VP of Product Operations and Chief of Staff to the CPO at Treasure AI
Treasure AI’s Kaori Ikeda Chun on product design with AI benefits
How do you build a steering wheel for something that’s already driving? For most of software’s history, the interface was where the work happened — you clicked, the thing did what you clicked. In an agentic product, a lot of the work happens whether you’re looking or not. So the interface stops being the place you do the work and becomes the place you understand it, trust it, and decide when to intervene. That’s a different craft. This week we’re running a four-part series with the product team at Treasure AI. Today we’re in the design seat — where all of the strategy finally becomes something a marketer can actually touch. To help me get into it, I’d like to welcome Kaori Ikeda Chun, VP of Product and Visual Design at Treasure AI.
FAQ
What is an agentic experience platform, and how is it different from a CDP? Flores describes it as the CDP retained underneath — centralized data, segments, audiences — with a decisioning layer and an activation layer added on top and an orchestration product, Treasure AI Studio, coordinating end to end. The distinction is that the system acts on customer data rather than only storing and segmenting it.
Why did Treasure Data rebrand to Treasure AI? The rebrand took effect April 20, 2026, after close to fifteen years under the Treasure Data name. Flores frames it as the same kind of category shift the company made around 2013, when data management and analytics became the CDP space: customers wanted AI decisioning and activation, and “CDP” no longer described the product.
Does Databricks entering the CDP market change Treasure AI’s strategy? Flores says it validates the direction and introduces a new competitor at the same time. Because customers already run Treasure AI alongside Databricks and Snowflake, the companies have to work together regardless. His competitive view is that warehouse-native entrants must still solve real time, which he identifies as the long-standing crux for data warehouses.
How long does it take to deploy a CDP now? Flores says a use case can be live in about two weeks, versus the months or years it historically took, with policy alignment and governance rather than data cleanup as the pacing constraint.
How should a marketing leader evaluate an “agentic” vendor claim? Flores’ test is immediacy: ask to see it and touch it on the spot. If the vendor needs to assemble a demo, or shows a prototype on render, they don’t have the product.
How many episodes are in this series, and where can I hear them? Four. Part one, with Rafael Flores, ran on the main podcast feed and is available on Apple Podcasts, Spotify, and YouTube. Parts two through four are video conversations embedded on this page and available on The Agile Brand YouTube channel.
Resources
Check out the Treasure AI Studio sandbox for yourself.
Enjoyed the show? Tell us more at and give us a rating so others can find the show.
Don’t miss a thing: get the latest episodes, sign up for our newsletter and more.
Check out The Agile Brand Guide website with articles, insights, and Martechipedia, the wiki for marketing technology.
