๐Ÿ’ซ PartnerPulse // CCoE

Real-time insights from partner interactions

โ˜๏ธ Cloud
๐ŸŽค Mic Only
๐ŸŽค+๐ŸŒ Mic + Tab
Initializing...
Press Start Transcribing to begin transcription

Transcript Input

Paste or type a transcript below, or use a transcribed session from the Transcribe tab.

Choose summary style

๐Ÿ’ฌ Ask PartnerPulse

Ask questions about your meeting summaries โ€” action items, decisions, trends, anything.

Hi! Select a partner above and ask me anything about your meeting summaries. For example: "What are the open action items?" or "Summarize the migration discussions"

Saved Summaries

No saved summaries yet
โ† Back to list

๐Ÿ“ค Sync Summaries
๐Ÿ“Ž Upload
๐Ÿ“‚ Browse KB
๐Ÿ” Query KB
๐Ÿ”’ Synced content is stored under the partner prefix and accessible only by the dedicated partner team. Ensure quality, meaningful data.
Loading saved summaries...
๐ŸŒ Upload documents that contribute meaningful context to the Global Partnership team. Files are stored under the selected partner prefix. Supported: PDF, DOCX, PPTX, TXT, MD, CSV, XLSX (max 25 MB).
๐Ÿ“„
Click to browse or drag & drop a file
PDF, DOCX, PPTX, TXT, MD, CSV, XLSX โ€” max 25 MB
๐Ÿ“„
Uploading...
Loading...
๐Ÿ” Ask questions across all synced meeting summaries and uploaded documents for a partner. Powered by Bedrock Knowledge Base with semantic search.
Hi! Select a partner and ask me anything about the Knowledge Base. I'll search across all synced summaries and uploaded documents from the entire team.

Try these example queries:
๐Ÿค–

Amazon Q โ€” My Assistant

๐Ÿ’ก How Amazon Q complements PartnerPulse:

QuickSight Q (My Assistant) gives you the quantitative view of your partner โ€” revenue, pipeline, ARR, deal stages, and win rates from structured data sources like Marshal and CRM.

PartnerPulse gives you the qualitative view โ€” meeting insights, action items, blockers, and relationship context from your conversations.

Use them together for a complete picture:

  • ๐Ÿ“Š Ask QuickSight Q: "What's the trailing 12-month revenue for HCLTech?" โ€” get the numbers
  • ๐Ÿ“š Ask PartnerPulse KB: "What are the key themes and blockers from recent HCLTech meetings?" โ€” get the narrative
  • ๐Ÿ”— Combine both to prepare for QBRs, write SIFT entries with real ARR data, or brief leadership with data-backed insights

Click below to open QuickSight and use My Assistant alongside PartnerPulse.

๐Ÿค– Open Amazon Q in New Tab
๐Ÿ“‹ QTR Tracker
๐Ÿ“Š SIFT
๐Ÿ“‹ Track Quarterly Technical Reviews across all partners. Admins can schedule QTRs, assign execs, and monitor completion. STLs can edit their assigned partners. Leaders have view-only access to the full portfolio. Summaries tagged as "QTR" auto-update the tracker.
PartnerNext QTRLast QTRExecLT MemberSTLStatusSummaries
Loading...
๐Ÿ“Š

Sales Insights & Field Trends (SIFT)

๐Ÿ’ก How to use PartnerPulse to draft SIFT entries:

  1. Go to the ๐Ÿ“š Knowledge Base tab โ†’ Query KB
  2. Select your partner and ask: "What are the potential SIFT entries that can be created from this KB?"
  3. Review the AI-generated SIFT entries โ€” each follows the SIFT template (Situation, Timeline, Actions, Impact, Next Steps, Ask)
  4. Pick the entry you want, copy it, and refine with specific dates, ARR figures, and named owners
  5. Click the button below to open SIFT and paste your entry directly

The Knowledge Base has the SIFT Guide indexed, so the AI knows how to structure Highlights, Lowlights, Observations, Blockers, Challenges, and Risks from your meeting summaries.

๐Ÿ“Š Open SIFT in New Tab
๐Ÿ“‹ Requests
๐Ÿ‘ฅ Users
๐Ÿข Partners
๐Ÿ“Š Stats
๐Ÿ’ฌ Feedback

๐Ÿ”ง User Access Requests

Loading...

๐Ÿ‘ฅ Current Users

All users with a profile in the system and their assigned partners.

Loading...

๐Ÿข Manage Partners

Add or remove partners. Changes apply to all dropdowns and signup forms instantly.

Loading stats...

๐Ÿ’ฌ User Feedback & Feature Requests

Click a sub-tab to load feedback
๐Ÿ’ซ Overview
๐Ÿ“– How-To Guide
๐Ÿ—๏ธ Architecture

๐Ÿ’ซ PartnerPulse // CCoE โ€” Capturing the Last Mile

Every partner meeting generates critical insights โ€” action items, technical decisions, migration blockers, commercial commitments โ€” but most of it lives in someone's head or scattered notes. PartnerPulse solves the "last mile" data capture problem by turning live partner conversations into structured, searchable, team-wide intelligence.

The problem: Partner-facing teams conduct dozens of meetings weekly. The insights from these conversations โ€” the real pulse of the partnership โ€” rarely make it into systems like SIFT or Amazon Q in a timely, structured way. Knowledge stays siloed with individuals.

๐Ÿš€ Key Capabilities

๐ŸŽค
Real-Time Transcription
Transcribe meetings directly in the browser (mic or mic + tab audio) with real-time speech-to-text powered by AWS Transcribe. Pause and resume as needed.
๐Ÿ”ฎ
AI-Powered Summarization
Summarize transcripts instantly using Amazon Bedrock (Claude) with multiple styles โ€” executive briefs, action items, technical deep-dives, TL;DR, or custom prompts.
๐Ÿท๏ธ
Partner Tagging & Isolation
Tag every summary to a specific partner. Each user is scoped to their assigned partner(s) โ€” dropdowns, KB stats, summaries, and queries are all filtered accordingly.
๐Ÿ’ฌ
Cross-Meeting Insights
Ask PartnerPulse questions across all your saved summaries โ€” spot trends, track action items, and prepare for follow-ups with multi-turn conversation.
๐Ÿ“š
Shared Knowledge Base
Sync summaries and upload documents (PDFs, decks, spreadsheets) into a partner-prefixed Knowledge Base. Only users tagged to that specific partner can query its knowledge base conversationally with source citations, maintaining strict partner isolation.
๐Ÿ”
Knowledge Base Querying
Semantic search across all synced content using Bedrock Knowledge Base + OpenSearch Serverless. Get answers grounded in actual meeting data with source citations. Shared guidance documents (like the SIFT Guide) are automatically included in all partner queries.
๐Ÿค–
Amazon Q & SIFT Integration
One-click access to Amazon Q (AI-powered analytics) and Salesforce SIFT โ€” connecting PartnerPulse to the broader analytics ecosystem.
๐Ÿ“
SIFT Entry Generation
Ask PartnerPulse to generate SIFT entries from your meeting summaries. The SIFT Guide is indexed in the Knowledge Base, so the AI structures responses using the proper SIFT template (Situation, Timeline, Actions, Escalation, Status, Impact, Next Steps, Ask) and categorizes them as Highlights, Lowlights, Observations, Blockers, Challenges, or Risks.
๐Ÿ”„
Auto-Indexing
Knowledge Base content is automatically re-indexed whenever summaries are synced or removed. New content becomes searchable within seconds โ€” no manual ingestion needed.

๐ŸŽค Transcribe Tab

  1. Choose audio mode:
    • ๐ŸŽค Mic Only โ€” Use for in-person meetings or when you only need to capture your own voice
    • ๐ŸŽค+๐ŸŒ Mic + Tab โ€” Use for virtual calls (Chime, Teams, Zoom, Google Meet) to capture both your mic and the remote participants' audio
  2. Click Start Transcribing โ€” you'll be prompted for audio consent and mic permissions on first use
  3. When using Mic + Tab mode, a browser sharing dialog will appear:
    • Select "A Tab" โ€” if your meeting is in a browser tab (e.g., Amazon Chime web app, Google Meet). Pick the tab running the meeting. Chrome will share that tab's audio automatically
    • Select "A Window" โ€” if your meeting is in a desktop app (e.g., Teams app, Zoom app). Pick the app window, then check "Also share system audio" at the bottom โ€” this is required to capture the remote participants
Choose what to share
A Tab A Window Entire Screen
๐Ÿ“น
Microsoft Teams
Application window
๐Ÿ’ฌ
Zoom Meeting
Application window
๐ŸŒ
Amazon Chime
Application window
โœ“
Also share system audio โ† REQUIRED
Cancel Share
๐Ÿ’ก Tip: For browser-based meetings (Chime web, Google Meet), "A Tab" is easiest โ€” audio sharing is automatic. For desktop apps (Teams, Zoom), use "A Window" and remember to check "Also share system audio".
๐ŸŽง Audio device note: Make sure your system audio output is set to the correct device before starting the recording. If you're using a headset (e.g., Jabra, AirPods, USB headset), the meeting audio routes through that device. Your OS audio output setting must match โ€” otherwise "Also share system audio" will capture silence. Check System Preferences โ†’ Sound โ†’ Output (macOS) or Settings โ†’ Sound โ†’ Output (Windows) and confirm the active device is the one playing the meeting audio.
  1. Speak naturally. Transcription appears in real-time with timestamps
  2. Use Pause to temporarily stop capture (e.g., during a break) without losing your session
  3. Click Stop Transcribing โ€” the transcript auto-saves to your account
  4. Use Copy to grab the full transcript, or head to the Summarize tab
๐ŸŒ Browser Compatibility
FeatureChrome / EdgeFirefox / Safari Login, Summarize, KB, Adminโœ… Full supportโœ… Full support Mic-only transcriptionโœ… Worksโœ… Works Mic + Tab/Window audioโœ… WorksโŒ Not supported
For the full transcription experience (capturing both your mic and remote participants), use Google Chrome or Microsoft Edge.

๐Ÿ”ฎ Summarize Tab

  1. Your transcript auto-populates, or paste any text manually. Click Use Transcript to pull from the Transcribe tab
  2. Pick a summary style: Executive, TL;DR, Actions, Technical, Detailed, Casual, or Custom
  3. Click Generate Summary โ€” Bedrock processes it in seconds
  4. Edit the summary inline if needed, then Save with a topic name and partner tag:
    • Partner summary: Select a partner name (e.g., HCLTech, Wipro) from the dropdown. Partner summaries can be synced to the shared Knowledge Base and are visible in team analytics.
    • Personal summary: Select ๐Ÿ”’ Personal from the dropdown. Personal summaries are private to you โ€” they are never synced to the shared KB, never visible to other users or admins, and only queryable by you via Ask PartnerPulse.
  5. Export as RTF for offline sharing
  6. Scroll down to Ask PartnerPulse to chat across all your saved summaries
  7. Saved summaries appear at the bottom, filterable by partner

๐Ÿ“š Knowledge Base Tab

  1. Sync Summaries: Select saved summaries and sync them to the shared S3-backed Knowledge Base. Use "No Sync" to hide irrelevant ones
  2. Upload: Drag and drop or browse for documents (PDF, DOCX, PPTX, TXT, MD, CSV, XLSX โ€” max 25 MB). Files are stored under the selected partner prefix
  3. Browse KB: View all synced items across the team, filtered by partner. Remove items if needed
  4. Query KB: Ask natural-language questions across all synced content for a partner. Supports multi-turn conversation and shows source citations. Try asking for SIFT entries โ€” e.g., "Generate a SIFT Highlight from the last HCLTech meeting"
  5. Auto-Indexing: Content is automatically re-indexed after every sync or delete โ€” new summaries become searchable within seconds

๐Ÿค– Amazon Q Tab

QuickSight Q (My Assistant) gives you the quantitative view of your partner โ€” revenue, pipeline, ARR, deal stages, and win rates from structured data sources like Marshal and CRM.

PartnerPulse gives you the qualitative view โ€” meeting insights, action items, blockers, and relationship context from your conversations.

Use them together for a complete picture:

  1. Ask QuickSight Q: "What's the trailing 12-month revenue for HCLTech?" โ€” get the numbers
  2. Ask PartnerPulse KB: "What are the key themes and blockers from recent HCLTech meetings?" โ€” get the narrative
  3. Combine both to prepare for QBRs, write SIFT entries with real ARR data, or brief leadership with data-backed insights

๐Ÿ”ง Admin Tab (admin users only)

  1. Requests: Review, approve, or deny access requests. Assign role (user/stl/leader/admin) and partner(s) during approval. Roles: User = standard access, STL = can edit QTR records for assigned partners, Leader = full app access + read-only QTR Tracker (portfolio oversight), Admin = full access including user management
  2. Users: View all current users, their email, role, and assigned partners. Search by email or partner
  3. Partners: Add or remove partners from the global list. Changes propagate to all dropdowns instantly
  4. Feedback: View all user-submitted feedback, feature requests, and bug reports. Filter by type or status. Update status (New โ†’ Reviewed โ†’ Planned โ†’ Completed โ†’ Declined) to track progress

๐Ÿ“‹ Governance Tab

The Governance tab contains two sub-tabs: SIFT (default) and QTR Tracker.

  1. QTR Tracker: Dashboard view of all partners with next/last QTR dates, assigned exec, LT member, STL, and status (Scheduled / Completed / Overdue / Not Scheduled). Summaries tagged as "QTR" auto-update the tracker status and date
  2. Scheduling: Admins and STLs can schedule QTRs using the "+ Schedule QTR" button. Assign exec, LT member, STL, dates, and notes. STLs can only edit partners assigned to them
  3. QTR Doc Upload: Attach QTR decks (PDF, DOCX, PPTX) directly from the edit modal. Files upload to the Knowledge Base and become searchable via KB Query
  4. Summary View: Click the ๐Ÿ“„ View or ๐Ÿ“Ž Doc badge in the Summaries column to see the QTR summary text or attached document details
  5. Role Permissions: Admin = full edit + schedule for all partners. STL = schedule + edit for assigned partners. Leader = read-only view of the full portfolio. User = view only, no edit buttons
  6. SIFT: Quick access to Salesforce SIFT for partner intelligence entries

๐Ÿ“ง Email Agent โ€” Outlook to Knowledge Base

The PartnerPulse Email Agent lets you search your Outlook inbox and send emails directly to the Knowledge Base โ€” from your own machine. Emails become searchable via "Ask PartnerPulse" within minutes.

  1. Prerequisites: Python 3.9+, Node.js 22+, aws-outlook-mcp (via toolbox install aws-outlook-mcp), and a valid Midway session (mwinit -s)
  2. Install Node 22 (Mac): brew install node@22 then add to PATH: echo 'export PATH="/opt/homebrew/opt/node@22/bin:$PATH"' >> ~/.zshrc && source ~/.zshrc
  3. Run the agent: python3 pp_agent.py โ€” opens a web UI at http://localhost:5555
  4. Sign in with your PartnerPulse credentials (same as this app)
  5. Search: Enter a topic (e.g., "HCLTech EBA"), select a partner, set days back, click Search
  6. Send to KB: Check the emails you want, add optional tags, click "Send Selected to KB"
  7. Verify: Go to Knowledge Base โ†’ Ask PartnerPulse and query about the email topic
How it works: The agent uses the same Outlook MCP server as Kiro, authenticated via your Midway session. Emails are converted to markdown, uploaded to S3, and indexed by Bedrock KB. The agent is read-only โ€” it never modifies or sends emails from your Outlook.
Get started: ๐Ÿ“ฅ Download Setup Guide โ€” Contact your PartnerPulse admin to receive the pp_agent.py agent file.

๐Ÿ’ฌ Feedback

We want to hear from you. Use the floating ๐Ÿ’ฌ Feedback button (bottom-left corner) to share your thoughts anytime.

  1. Click the ๐Ÿ’ฌ Feedback button โ€” a modal will appear
  2. Choose a feedback type:
    • ๐Ÿ’ก Feature Request โ€” Suggest a new capability or improvement. You can set priority (must-have, nice-to-have, or low)
    • ๐Ÿ’ฌ General Feedback โ€” Share thoughts, impressions, or ideas about the app
    • ๐Ÿ› Bug Report โ€” Something broken or not working as expected? Let us know
  3. Write your message and click Submit
  4. Your feedback is stored and visible to admins, who will review and track it
๐Ÿ’ก Tip: Feature requests help us prioritize what to build next. The more specific you are, the better we can act on it. Don't hold back โ€” every idea counts.

๐Ÿ—๏ธ Architecture Overview

PartnerPulse is a serverless application built entirely on AWS managed services. No servers to manage, scales automatically, and pay-per-use.

๐Ÿ–ฅ๏ธ Presentation Layer
โ˜๏ธ CloudFront CDN ๐Ÿ“ฆ S3 Static Hosting ๐ŸŒ Single-Page App (HTML/JS)
Browser tabs: Transcribe ยท Summarize ยท KB ยท Amazon Q ยท SIFT ยท Admin ยท Help
โ–ผ HTTPS
๐Ÿ” Authentication Layer
๐Ÿ”‘ Federate OIDC + Cognito ๐Ÿ”„ PKCE Auth Code Flow ๐Ÿ‘ฅ Roles: admin / leader / stl / user
No self-registration ยท @amazon.com / @amazon.co.uk only ยท Corporate identity ยท Lambda@Edge validation
โ–ผ Bearer Token
๐ŸŒ API Layer
๐Ÿšช API Gateway (REST) ๐Ÿ”Œ API Gateway (WebSocket) ๐Ÿ›ก๏ธ Lambda Authorizer
REST: /transcripts ยท /summaries ยท /admin/* ยท /profile ยท /config ยท /knowledge-base/*
WebSocket: summarize ยท kb-query ยท insights (no timeout limit โ€” persistent connection)
โ–ผ Lambda Proxy Integration
โšก Compute Layer
CRUD Lambda
Node.js 22 ยท Profiles, Summaries, Transcripts, Admin, KB Query (REST fallback)
WebSocket Lambda
Python 3.12 ยท JWT auth on connect ยท Routes: summarize, kb-query, insights ยท 120s timeout ยท Auto-reconnect
Summarize Lambda
Bedrock Claude Sonnet 4.5 (all styles) + Guardrails ยท Content filtering
KB Sync Lambda
S3 sync + Bedrock Data Source ingestion
Access Request Lambda
Public signup endpoint ยท Admin approval workflow
Transcribe URL Lambda
Pre-signed WebSocket URL generation
โ–ผ
๐Ÿง  AI / ML Layer
Amazon Bedrock
Claude Sonnet 4.5 (all styles) ยท Summarization ยท Cross-meeting insights ยท KB Query ยท SIFT entry generation ยท Guardrails (PII + content filtering)
Amazon Transcribe
Real-time streaming STT ยท Pre-signed WebSocket ยท 16kHz PCM
Bedrock Knowledge Base
OpenSearch Serverless vectors ยท Semantic search ยท RetrieveAndGenerate ยท Source citations
โ–ผ
๐Ÿ’พ Data Layer
DynamoDB
User profiles ยท Summaries ยท Transcripts ยท Access requests ยท Partner config ยท WebSocket connections (TTL) ยท Encrypted at rest
S3
Knowledge Base docs (partner-prefixed) ยท Frontend assets ยท Server-side encryption
OpenSearch Serverless
Vector embeddings collection ยท Automatic scaling
๐Ÿ›ก๏ธ Security & Governance
๐Ÿ”’ HTTPS everywhere ๐Ÿ”‘ OAuth2 PKCE ๐Ÿ›ก๏ธ Bedrock Guardrails ๐Ÿ‘ค Partner-scoped isolation ๐Ÿšซ No self-registration ๐Ÿ“ง @amazon.com / @amazon.co.uk ๐Ÿ” Encryption at rest ๐Ÿ“‹ CloudWatch logging ๐ŸŒ WAF Protection ๐Ÿ”— CORS Locked โšก API Throttling

๐Ÿ“ AWS Reference Architecture

Full architecture diagram with official AWS service icons โ€” click to view full size.

PartnerPulse Cloud Architecture โ€” AWS Reference Diagram

๐Ÿ”’ Security & Privacy Details

  • Authentication: Amazon Federate OIDC via Cognito User Pool โ€” corporate identity authentication with PKCE authorization code flow. No separate passwords required
  • Authorization: Role-based access control (admin/leader/stl/user) from DynamoDB PROFILE records, enforced at API Gateway (Lambda authorizer) and Lambda level
  • Partner isolation: User profiles store partner assignments. All dropdowns, queries, KB stats, and data views are scoped accordingly
  • No self-registration: Access is request-only, reviewed and approved by admins. Only @amazon.com and @amazon.co.uk emails accepted
  • Data at rest: DynamoDB encryption (AWS-managed keys), S3 server-side encryption
  • Data in transit: HTTPS/WSS everywhere โ€” CloudFront โ†’ API Gateway (REST + WebSocket) โ†’ Lambda. Transcribe WebSocket uses pre-signed URLs with short TTL. AI endpoints use persistent WebSocket with JWT auth on connect
  • Bedrock Guardrails: Content filtering applied to summarization to prevent harmful content generation and filter sensitive information
  • Observability: Client-side errors captured and sent to CloudWatch via /logs endpoint
  • WAF Protection: CloudFront protected by AWS WAF with managed rule sets (Common Rule Set, Known Bad Inputs) and IP-based rate limiting (1000 requests per 5 minutes per IP)
  • CORS Lockdown: All Lambda functions restrict Access-Control-Allow-Origin to the CloudFront domain only โ€” no wildcard origins allowed
  • API Throttling: API Gateway stage-level throttling at 50 rps / 100 burst. Access-request endpoint further limited to 5 rps / 10 burst to prevent abuse