DataStax
2025
Shifting to an AI-Ready Database
Streamlined the database creation process while introducing the new Vector deployment type for GenAI applications.
The Context
A product at a strategic inflection point.
Astra DB had been built around two deployment models — Serverless and Managed Clusters — serving distinct personas with very different needs. As the AI/ML wave accelerated, DataStax made a strategic bet: launch a third type, Serverless (Vector), built for GenAI workloads.
The business challenge was adoption of this new deployment type. The existing creation flow treated all deployment types equally, burying the new Vector type inside an already confusing experience. Without a deliberate design strategy, the new deployment type risked becoming just another item in a dropdown.
The Current State
Redesigning a high-friction database creation journey.
I analyzed the existing user journey to resolve usability bottlenecks and integrate new deployment types. My strategy focused on understanding the current pain points and what we could improve in elements such as wording, layout, and plan selection.
The Insights
Developers didn't speak "Cassandra".
We interviewed 10 ML engineers to understand their mental models. The finding that shaped everything: these users cared about one thing, whether the database would work with their AI stack. References to “C*” and cloud-provider specifics created friction before users even reached the creation step.
Removing the technical jargon
8 out of 10 engineers found the naming and system labels unclear.
Lack of cloud provider context
7 out of 10 engineers were uncertain about provider selection.
Preference for a GenAI type
9 out of 10 engineers selected Serverless options over standard Managed Clusters.
The Iterations
Three approaches, one real tradeoff.
We explored three distinct directions. The core tension was how aggressively to push the business strategy through the UI: driving adoption of Vector versus preserving easy access for enterprise buyers.
01
Segmented pathways
Separate flows for enterprise vs. serverless users. Tailored but fragmented, this direction risked creating two separate flows and obscuring the new offering from enterprise users who might benefit from it.
02
Unified dialog
Single flow, with Serverless (Vector) pre-selected as the default. Managed Clusters remain accessible but deprioritized. Drives adoption through framing, not restriction.
03
Paywall gating
Managed Clusters visible but locked behind payment input. Serverless (Vector) remains the frictionless default. Managed Clusters carry significant cost implications, and accidental provisioning is a serious problem for enterprise buyers. The friction is deliberate, proportional, and serves the user as much as the business.
The Solution
A modular, AI-first creation flow.
The approved design unified all deployment types in a single dialog, with Serverless (Vector) pre-selected as the clear default. Managed Clusters remained available behind a deliberate gate, shifting the cognitive default toward AI workloads without removing enterprise optionality.
The creation flow was built as a modular framework, designed to absorb future deployment types without accumulating UX debt.
The Impact
Strategy validated by adoption.
DataStax became a market leader according to Forrester. We drove growth to $70M ARR, and Serverless (Vector) became the dominant choice for new databases, validating our AI-first deployment model.
