Pulse - AI Agent Performance Monitoring and Analytics for Amazon Connect

Pulse is an AI performance management platform for self-service AI agents in Amazon Connect. It reads every chat and voice conversation your AI agents have, groups them by the reason the customer got in touch, and gives operations, CX, quality and engineering teams one shared view of what your customers experience. Its Insights Center ranks every topic by how well your AI is resolving it and how that is changing. Pulse works with the self-service AI agents you already run; AWS remains the agent foundation. Teams use it to observe performance, investigate causes, detect emerging issues and tie results to customer outcomes.

Outcomes you can see

• Know which topics your AI resolves, which it loses to a live agent, and where customers give up, and how that changes week over week and month over month
• Find the tool, knowledge source or step behind a poor result without reading transcripts one by one
• Control who can see customer conversations, with custom roles and PII redaction

Detect: Insights Center

• Spot which topics need attention before the top-line KPIs move: see each topic by session volume and resolution rate against your benchmark, sorted by unresolved sessions, predicted satisfaction or resolution decline
• Compare any period with the one before it, such as week over week or month over month, to see which topics are improving or declining; new topics are shown separately until they have enough history to compare fairly
• Open a topic to see resolution rate, predicted satisfaction, escalation to a person, repeat contact and repeated requests, each compared with all other topics
• See the failure signals behind it: tool call failures, knowledge gaps, prompt failures, response time and tokens per session
• Break a topic down by failure cause, tool, knowledge source, agent or prompt version, or channel, open the matching sessions in Session Logs, and share a link to the view

Observe: Command Center

• 30+ KPIs across business, operational and diagnostic views, scoped by AI agent, channel (chat or voice), region and time period, with trend views by hour, day or week
• Headline business measures: self-service resolution, goal success rate, escalation rate, predicted customer satisfaction (scored on every conversation, not only surveyed ones) and abandonment
• Open the sessions behind any metric and export any chart to CSV

Investigate: Session Logs and Trace Explorer

• Search transcripts and combine up to 10 filter conditions, including failed or empty AI responses and customers who asked for a person
• Read each conversation with the tools and APIs called, knowledge sources retrieved, latency and token use per turn
• Save and share filters, and export the filtered list to CSV with the columns you choose
• View Amazon Connect Contact Lens sentiment and redaction beside the conversation, when Contact Lens is configured

Evaluate Quality

• Define evaluation forms in Amazon Connect; Pulse brings scores, reasoning and evidence into Command Center and Session Logs, where you can filter by result

Topic Management

• Create topics in your own words and assign them to AI agents; Pulse checks your topics first, then falls back to an AI-generated topic
• Promote AI-generated topics, merge overlapping ones (merged names are kept as aliases), and choose which topics appear in reporting with Include in Reporting

Security and Administration

• Custom roles with View, Create, Edit, Delete and Decrypt permissions; Decrypt is a separate, limited permission
• Control how session attributes are retained and encrypted, and redact personal data at deployment. Redaction reduces risk but may not catch every instance
• Deployed by NeuraFlash into your own AWS account and Region

Optional Feature: Connect AI Chat Experience

• Add branded chat, reusable rich chat templates (quick replies, carousels, forms, time pickers) and a shared asset library to Amazon Connect chat, without changing the AWS agent underneath