Alexa+ Developer Agent
A conversational AI that guides third-party developers through building smart home integrations for Alexa+
Role
UX Designer
Industry
Developer/AI Tools
Duration
6+ months

The problem and opportunity
Alexa’s standard APIs couldn’t accommodate the custom functionality growing in the smart home device market. Third-party developers had no guided path: they navigated complex schema requirements without support, tested through email exchanges with the Alexa team, and waited months to deploy a single integration. The opportunity was to open Alexa+ to a broader ecosystem of hardware developers while making the process dramatically faster.
Research
Structured discovery interviews: 12 developers and internal stakeholders, discussion guide, uncovered 4 systemic problems
Fragmented process: creation, testing, and UI lived in disconnected tools with no relationship between them
No testing visibility: testing happened over email, developers had no insight into failures or how to fix them
No guidance: developers relied on third-party docs to understand Alexa’s schema requirements
Months to deployment: disconnected tools and lack of guidance made a single integration take months
Design
Journey mapping: mapped the full E2E developer journey across 5 stages to align design, product, and engineering before any screens were designed
Designing the agent: wrote a system prompt to define the agent’s conversational approach and behavior, then translated every schema concept into plain-language questions a developer could answer without technical knowledge
AI-assisted prototyping: used Kiro to generate wireframe flows and interactive prototypes to align product and engineering early, integrated a design system into prototypes to accelerate handoff and build
Live prototype testing: connected prototypes to Amazon Bedrock and Claude APIs to test with developers using real device use cases, enabling fast iteration grounded in actual agent behavior

Solution
A conversational AI console that guides developers through the full integration journey — describing their device in plain language, building the schema in real time, testing, debugging, and deploying. Developers never interact with raw schema. Advanced developers can upload existing documentation and the agent fills gaps through targeted questions.


Outcomes
• Under 30 min to a first draft integration, down from weeks of manual schema work
• Tareting 70% first-submission pass rate targeted at launch, up from an estimated 30-40% today
• Targeting 50 net new integrations projected in the first 90 days post-launch