Jabra Engage AI - Jabra Engage AI

Jabra Engage AITechnical OverviewJabra Engage AI – Technical OverviewContents

Thank you for reading Jabra Engage AI – Technical Overview

Description

AI Technology

Architecture

Software Operation

System Requirements

Deployment

Data Handling

Description

Jabra Engage AI is a tool for contact centers to deliver outstanding customer experience, powered by tone of voice AI. Every agent is presented with real-time feedback for self-coaching. Every supervisor sees the real-time view of their teams.

We believe the big potential in AI is to empower people, not replace them. Engage AI allows you to understand conversations in a completely new way, through tone of voice and conversation metrics. It increases agent engagement, brings teams closer and works in complete harmony with your preferred contact center platform. Read more at https://jabra.com/engageAI.

AI Technology

Jabra has partnered and invested in audEERING, the world’s leading innovator in voice AI. audEERING roots trace back to the AI research group by Prof. Björn Schuller (TUM). Together with Dagmar Schuller, Dr. Florian Eyben, Dr. Felix Weninger and Dr. Martin Wöllmer, the company was founded in 2012.

This internationally renowned research team has over 20 years of experience in Audio AI. audEERING is owner of the open-source audio analysis toolkit openSMILE.

Architecture

Jabra Engage AI is a hybrid solution: a client application installed locally on the PC and a cloud backend.

The key connection is between the softphone and Engage AI. It is critical for Engage AI to receive call control events (e.g. call start/stop/pause/transfer/…) in order to function correctly.

The client application serves to:

Listen for softphone commands to start/stop analysis
Perform the analysis on the audio stream
Display results to the user

The cloud backend serves to:

Populate dashboards
Provide access to the data API
Manage configuration and settings

The cloud service is operated by Jabra in a multi-tenant architecture. The cloud is hosted by Microsoft Azure in the EU region or in Deutsche Telekom’s OTC.

Software Operation

The software runs in the background on the PC. When it receives a call start signal from the softphone, analysis begins. The call audio is taken from the headset and processed through the AI model. Results are shown on-screen for the user. Statistics and metadata are sent to the cloud. Analysis is terminated when the call stop signal is received from the softphone.

The audio stream is only processed on the local PC. The audio stream is never moved to the cloud. All data is encrypted in transit and at rest.

Each user has their own account. This can be a username/password combination or can rely on Azure AD SSO.

System Requirements

The software has requirements on the local PC and on network connectivity.

CPU	Intel i5 or similar
RAM (free)	600 MB typical, 800 MB peak
Disk (free)	700 MB
Network	Internet access required, 10 Kbps average consumption
Operation System	Windows 10, 11
Note on VDI	VDI environments are supported with additional testing
Softphone	Standard connectors are available for certain softphones (e.g. Genesys Purecloud). Custom connectors are possible.

All the communication from the Jabra Engage AI client happens on HTTP(s), and uses standard ports 80 & 443. Communication is required to certain endpoints as described in the “Firewall Rules” section.

Deployment

The installer is typically provided for customer IT to deploy. The software is automatically updated from Jabra’s central service.

Data Handling

Data can be separated into two broad categories: application data and technical logs.

The application data contains agent username, email, call metadata, and analysis results on call history. No caller data (name of person calling in, inbound phone number, ID numbers, etc.) is captured or stored. Application data is maintained for as long as the account is active plus 30 days, or as otherwise specified in the Master Subscription Agreement.

The technical logs contain logging information (IP address, OS version, etc.) and retained for 30 days.

Jabra Engage AI is fully GDPR compliant. For additional details on privacy, please contact JimWang@jabra.com

Supported Integrations & Architecture

Below you can find the overview of most used integrations. For the full list please visit: Supported softphones | Supported softphones (clickup.com)

The integration architecture shows how Engage AI integrates within a contact center environment. It analyzes calls 100% locally using built-in AI models. There are different ecosystems and scenarios with integrations are highlighted using yellow color.

Integration basics

To integrate Engage AI to analyze calls, following components are needed:

Call control events
Audio streams

There are different ways Engage AI could integrate call control events both locally and remotely, but audio streams are analyzed only locally on a PC or on a remote VDI desktop if audio is accessible.

For the local integration, call control events are considered in the scope of current user running Engage AI application instance. For the remote integration the agent email address is required. The email address must match the email address used by agent to login into the Engage AI application.

The remote integrations (e.g. WebApi) requires an ApiKey to be send in the HTTP request header. The ApiKey will be provided for customer by the Engage AI team.

Integration details

Local call control events integrations details

The preferred option for integration with Engage AI is local one as we are driven by privacy and real time audio processing

Command line

You can call Engage AI executable with following properties to provide call control events

Parameter	Value	Details
callcontrol	N/A	Required for call control events
<action>	N/A	Call control event. Available actions: callstart, callend, callpause, callresume
callKey	ID	Current call ID
optional_property	optional_property_value	Any key value pair

Example call:
engage.ai.exe --callcontrol --callstart --callkey:callId-1 --customProperty1:customValue1

Named Pipe

We can implement cross process communication using Windows Named Pipe technology.

https://learn.microsoft.com/en-us/windows/win32/ipc/named-pipes

Custom plugin or add-on

If your software supports plugins or add-ons, we can implement integration with Engage AI. The software integration documentation should be shared with the Engage AI team.

Remote call control events integrations

The remote option is also available but introduces slight delays.

Both Web Api integrations are secured with API key generated by Jabra Engage AI team for each customer.

Web Api for call events

Call events can be sent to the Web Api endpoint. Message format:

Parameter	Type	Details
userName	String	Email address for advisor
message	Object
message.state	String	Call control event. Available actions: callStarted, callEnded, callPaused, callResumed
message.callKey	String	Current call ID
message.properties	Dictionary	Key value pair dictionary

Example:
curl -X POST 'https://smile.jabra.com/v1/api/call-integration' -H 'x-api-key: <api-key>' -H 'Content-Type: application/json' -d '{\"message\":{\"state\": \"CallStarted\"}, \"userName\":\"<user email>\"}'

Web Api for call state

Call state can be send to our Web Api. Message format:

Parameter	Type	Details
userName	String	Email address for advisor
message	Object
message.state	Object
message.state.isActive	Boolean	Current state
message.state.isPaused	Boolean	Is call ongoing or paused
message.state.isMuted	Boolean	Is call muted
message.callKey	String	Current call ID
message.properties	Dictionary	Key value pair dictionary

Example:
curl -X POST 'https://smile.jabra.com/v1/api/call-integration/state' -H 'x-api-key: <api-key>' -H 'Content-Type: application/json' -d '{\"message\":{\"state\": {\"isActive\":true, \"isPaused\":false, \"isMuted\": false}‌}, \"userName\":\"<user email>\"}'

Custom

If you expose Web Api , Web Socket or Web Hooks for call control events we can implement integration with Engage AI. The software integration documentation should be shared with the Engage AI team.

Firewall Rules

Jabra Engage AI URIs:

For MS Azure version:
https://smile.jabra.com/ - Jabra Engage AI backend

For DT OTC version:
https://*.engageaix.jabra.com - Jabra Engage AI backend

For both MS Azure and DT OTC versions:
Jabra Engage AI uses GN Unified Identity Platform (UIP) for user authentication/authorization. GN UIP is based on Microsoft Azure AD B2C so its resources should be accessible also.

URIs:

https://account.gn.com - GN UIP front door
https://stognuipprod.blob.core.windows.net - static resources of GN UIP
https://stognuipprod.z6.web.core.windows.net - static resources of GN UIP
https://**.msecnd.net - static resources of Microsoft Azure AD B2C

Jabra Engage AI uses three industry-standard logging and diagnostics systems, one of them is Solarwind Loggly , another one is Microsoft Azure Application Insights and the third is Google Analytics.

URIs:

https://logs-01.loggly.com/ - Solarwind Loggly logs collection endpoint
https://dc.services.visualstudio.com/ - Microsoft Azure Application Insights logs collection endpoint
https://www.google-analytics.com/ - Google Analytics endpoint