top of page

Cross-Device Conversation Continuity

In 2025, Amazon launched Alexa+, a new version of Alexa powered by a Large Language Model. The team anticipated that customers would start using Alexa for longer planning discussions and content creation. The goal of Cross-Device Conversation Continuity was to let customers pick up where they left off, access generated artifacts, and transition smoothly between endpoints. 

My Role

  • Principal UX Designer

  • Design Lead

Working Group

  • Principal Product Manager

  • Sr Conversation Designer

  • Principal Engineers

  • Endpoint UX Designers

Stakeholders (200+)

  • SVP of Amazon Devices

  • Several Alexa VPs

  • Alexa Core Design and Product Directors

  • Domain Product, Design 
and Engineering teams

Customer Jobs To Be Done

Find and continue a prior conversation

Get back to artifacts generated by Alexa

Transition to the best endpoint for the job

Challenges

  • Customers in an early beta trial expected to be able to review full conversations and artifacts, not just one-off requests which often lacked full context. 

  • Competitive research showed that while other AI agents offered session histories, they lacked cross-thread context. 

  • Alexa is primarily accessed by voice on communal devices in the home rather than a GUI interface on a personal phone or computer. 

  • Many interactions with Alexa on Echo devices are less conversational and more transactional in nature (e.g. set a timer)

“…I want to be able to look at them later. In chat GPT you can go back to that. Maybe I don't need to write it down. Maybe she'd remember it. She had good activities for the kids. ... I'm not going to sit and write notes...”

Beta participant,
Current ChatGPT user

Our Continuity system needed to support a natural voice-forward way to resume a past chat and the ability to differentiate actual chats from basic requests.

But, how?

Insight & Strategy

We explored multiple models including Freeform memory (recall like human memory), Temporal sessions (chronological history) and Topic-based grouping (auto-organized conversations). We aligned on a hybrid model, combining chronological chat history with memory-based context.

Freeform Memory

Frame 47_edited.png

Temporal Sessions

Topic-based Groupings

Our Continuity system needed to support a natural voice-forward way to resume a past chat and the ability to differentiate actual chats from basic requests.

Memory + Temporal sessions gave customers clear, durable guideposts on screen-based endpoints and a natural way to initiate and resume chats by voice.

Execution

  • Authored a comprehensive specification covering conversation boundaries, logic for differentiating chats from basic requests, cross-device presentation, and profile handling. This living document was referenced by over 200 stakeholders and engineers.

  • Partnered with engineers to break the project into deliverable phases.

  • Advocated to keep resuming chats with voice in scope; despite technical complexity and invention risks.

  • Navigated leadership misalignment by escalating competing visions.

  • Aligned on a coherent, customer-focused solution and then created example flows and patterns to clarify intended behavior for endpoint design teams.

Example Scenario

Jackson needs dinner ideas

Jackson asks Alexa for some dinner recipe ideas.

Alexa provides Jackson ideas on his Kitchen Echo Show.​ 

Man talking to Echo Show

Results

Cross-device Conversation Continuity shipped to customers with the launch of Alexa+ in 2025 and received positive press coverage.
 

  • Exceeded beta goals: CSAT 6.25/7 and 92% accuracy for resuming conversations by voice.

  • Beta customers praised the ability to continue conversations seamlessly across devices and to review history without note-taking.

  • Filed a patent for the innovative hybrid design approach.

  • Established a strong foundation for future memory aggregation and long-term project management features.

bottom of page