Frequently asked questions

Jaypaper performs Voice Activity Detection (VAD) on the raw audio signal from your Amazon Connect calls. We detect speech presence, silence, crosstalk (both parties talking simultaneously), and timing patterns. We never transcribe or record the conversation — we only measure the acoustic characteristics of the signal.

No — it adds to them. Jaypaper analyses acoustic signals to understand how conversations behave — silence, interruptions, response timing, talk balance, crosstalk. It's behavioral and deterministic: every score is reproducible, and supervisors can tune the thresholds. It runs alongside your existing Amazon Connect analytics, adding an always-on behavioral layer on top of whatever content analysis you already use.

No. Audio is analysed in real-time and immediately discarded. We store only acoustic features — timing data, energy levels, speech presence indicators. The meaning of the call is never captured or stored. There are no transcripts and no recordings. Caller ID (phone numbers) is only stored if your organisation explicitly opts in.

Deploy our CloudFormation template in your AWS account (~5 minutes). It creates the resources Jaypaper needs to receive your call data — a Kinesis Data Stream for agent events, an EventBridge rule for contact lifecycle events, and a read-only IAM role. Then install our contact flow module in your Amazon Connect instance — a pre-built integration block that starts media streaming and connects to Jaypaper. No changes to agent desktops, routing, or how calls are handled.

Our CloudFormation template creates: a Kinesis Data Stream for agent events (~$11/month), an EventBridge rule that forwards contact lifecycle events to Jaypaper (free), and two IAM roles — one for EventBridge forwarding, one granting Jaypaper read-only access to your Connect instance (list queues, list agents). Audio flows through Kinesis Video Streams, which Amazon Connect provisions natively when media streaming is active. That streaming carries a small, usage-based AWS charge in your account on top of the agent-events stream — typically a few dollars a month, scaling with call volume; we're still finalising exact figures ahead of GA. You own everything and can delete the stack at any time.

By default, read-only. Jaypaper can list your queues, agents, and flows — nothing else. If you enable flow automation (optional), Jaypaper can also create and update contact flow modules on your behalf — this is how one-click setup works. Both levels of access are granted via an IAM role you control in your own account. Delete the CloudFormation stack and all access is immediately revoked. We never have credentials to your AWS account.

Through a live dashboard. You see all active calls with real-time behavioral scoring, can expand any call for a full metric breakdown and timeline, and filter by date range for historical analysis. Scores update live as conversations progress. Thresholds are tunable — define what "good" means for your team.

Response latency, interruptions, crosstalk percentage, monologues, silence events, agent talk time, greeting detection, and close detection. Each metric has a configurable weight and scoring type. The composite score (0-100) recomputes live on every VAD event. Supervisors can adjust thresholds and weights to match their team's standards.

5,000 minutes of acoustic analysis per month, permanently free — about 83 hours of call audio. No trial expiration, no credit card required. If you exceed your included minutes, acoustic analysis pauses but call state monitoring stays on — you still see active calls, queues, and agents. No overage charges, ever.

Acoustic analysis pauses: silence detection, crosstalk scoring, and talk pattern metrics stop processing new calls. Everything else continues — active call visibility, queue status, agent availability, call duration tracking, and historical data you've already accumulated. Think of it as losing the "how does this call sound?" layer while keeping the "who's on a call and where?" layer. Upgrade or wait for your period to reset, and full analysis resumes instantly.

We're starting with us-east-1 and expanding based on demand. The Kinesis stream in your account must be in a region we support. Let us know your region when you sign up for early access and we'll prioritize accordingly.

Not during beta. We're committed to best-effort uptime and will be transparent about any issues. A formal SLA will come with general availability.

Not automatically. Amazon Connect allows only one Agent Events destination and one Live Media Streaming destination per instance — our CloudFormation template creates and owns its own, rather than attaching to one you already have. If either is already configured for another integration, our setup will fail cleanly (the whole stack rolls back automatically) rather than silently overwrite or coexist with it. Email support@jaypaper.com before you deploy and we'll help you connect without disrupting what you already have.

Revoke the IAM role (or delete the CloudFormation stack) and Jaypaper immediately loses access to new call data. All stored acoustic features are deleted within 30 days of cancellation.