Dextr steps closer to AI in the Call Center with LENS integration!


Dextr integrates transcriptions, comprehension and sentiment analysis !

Contact center supervisors are hungry for insight into things like whether agents are having effective conversations, whether any interesting trends in customer sentiment are occurring, or whether the agents are complying with regulatory requirements. This is challenging because audio data is virtually impossible for computers to search and analyze. Therefore, recorded speech needs to be converted to text before it can be used in applications.  Using AWS machine learning natural language processing (NLP) and speech-to-text, Contact Lens for Amazon Connect, now integrated in the Dextr Dashboard,  transcribes contact center calls to create a fully searchable archive and surface valuable customer insights.  With Contact Lens for Amazon Connect, customer service supervisors can quickly and easily discover emerging themes and trends from customer conversations, directly in Amazon Connect.

Dextr simple to navigate interface

Dextr has long enabled agents and supervisors to search activities and pull back voice recordings.  We the integration of LENS, Dextr makes it possible to not only pull back the recording, but to “see” key customer experience metrics including total talk time in bar chart format showing customer, agent and silence time slots.  The sentiment analysis is displayed by segment and rated over all by a smiley face!

Transcriptions can be displayed by agent and customer segment, with clickable time stamps that enable you to pull back any segment of a recorded and transcribe customer experience.

Dextr Benefits

With a Dextr dashboard, contact Lens for Amazon Connect simple to set up and use. With only a few clicks, you can use Machine Language -powered analytics to discover deep customer insights from your contact center. Through the Dextr intuitive UI, you can analyze call and chat transcripts, customer and agent sentiment, and conversation characteristics without any coding.  The Dextr implementation of Contact Lens for Amazon Connect helps supervisors improve the agent customer interaction by providing insights around what makes a successful call, compliance risks, and trends in topics with new rich metadata around call transcripts and conversation characteristics such as indicators of interruptions, talk speed, sentiment, and custom category labels.

Dextr subscriptions are available in the AWS Marketplace

Dextr is a subscription based agent dashboard with a cost of about $1 a day per eight hour agent shift on a metered basis.  Dextr offers discounts for prepaid contract terms and offers ongoing support for your Amazon Connect Call Centers.  If you have not yet migrated to Amazon Connect, the Dextr professional service team provides a “fixed cost” deployment option that can have you operational in days and includes a one year subscription to Dextr.  For your free trial head over to AWS Market place and sign up now!


Advanced Post Call Survey using Amazon Connect and LENS!

Post Call Survey configurations

Configuring a post call survey in Amazon Connect is very straight forward and many others have outlined strategies for doing such a configuration.    The usual strategy typically prompts the caller to enter a “digit” from 1 to 5 to rate a particular agent or service.   Then, if the caller has not yet hung up, you prompt them with yet another question and so forth until you either exhaust the patience of the caller of tire of asking to many questions.

The answers are typically stored as contract attributes that become part of the contact record for that phone call.  Then an outside report generator would have to be created to read the data and provide a human readable summary of survey results.  Typically creating the administrative interface and reporting facilities take more time an effort that creating the post call survey!   The good news is this is a workable strategy and can get the job done, it is dirt cheap and easy to do!

Why not use Sentiment Analysis?

Why exhaust the caller with 20 questions?  There is a better way and it is more effective strategy that provides even more details about the customers experience while interacting with your call center team.   Using the same facilities that we described in our previous blog on voice mail transcription options, you can now use LENS to provide a deeper understanding of the customer experience.  Creating a post call survey using these tools enables you to “hear” the customers voice if you like,  transcribe the voice and use sentiment analysis to better grade both agent performance and customer satisfaction.

Enter Media Streaming!

In the following call flow, we can provide a post call question and enable the caller to freely speak their experience which is recorded. The recording can be free form, allowing the caller to use a stream of conscience approach to outline their experience.  This is powerful because you can, if you chose to, hear spoken nuances that easily identify frustration and other indicators of how people really feel about their call center experience.

If you are insistent and really want to ask individual questions, you can loop through steps 1 through 4 asking a different question at step 1 each time.  The entire recording is saved with a contact record ID that associates the recording with a specific contact trace record.  By setting the recording and analytics behavior when you set the recording, you now enable LENS to kick in and add that extra level of analysis that only comprehension and sentiment analysis can provide.

Let us know if we can provide any assistance in getting this setup in your connect instance.  Just open a ticket with  and we will get you “connected”. – DrVoIP








Direct Extension Dialing and queue based Voice Mail for Amazon Connect!

Call Center with PBX like features?

One of the most requested features folks deploying Amazon Connect typically ask for is the ability to directly dial an extension to reach a specific agent or team member.   Clearly Amazon Connect is not a PBX system, it is a call center!  It is possible, however,  to give your Call Center instance qualities that enable PBX like functions including voice mail, Extension dialing and even queue based voice mail.   In a world with geographically dispersed workgroups and most folks using cell phones anyway, a cloud based solution based on Amazon Connect is ideal, practical and easy to administer.

Dial by Name Option

Why we still use DTMF Call Trees in the 21st Century is a mystery to me.   AWS provides speech recognition technology that enables you to replace the old world “Press 1 for this and Press 2 for that” with “So glad you called, I can route your call if you tell me the name of a team member or a function like sales or customer service”.    This interface is an improved customer experience in several dimensions.  It enables “hands free” interactions for mobile users,  is more natural and experientially compatible to human communications.

Dynamic Configurations

Every resource in the Amazon Connect instance including prompts, queues, flows and even agents have a unique identification number.  This number, or ARN,  can be used in contact flows to check the status of a particular user for example.  They can also be used to dynamically configure a contact flow with end points that change on a call by call basis.  An incoming call, for example, can invoke a lambda function that is used to look up the ‘dialed number” in an external database and bring back queue details that can be used to dynamically populate the details of the contact flow.   Rather than create a separate contact flow for “sales”, “customer service and “technical support” why not use the same contact flow, reconfigured based on either the callers self navigation choice or the DNIS the caller dialed?   It takes nothing more than a simple dynamoDB table to quickly define the various required flow parameters and an “invoke lambda” function to greatly simplify the other wise complex world of “cut and paste” contact flows.

Agent ID

Each Agent in a call center instance has an ARN that uniquely identifies that user.  This ARN when defined in an external database that also specifies an extension number, cell phone, user name and even a custom prompt can be configured to create very powerful contact flows.  A caller can either ask LEX for a particular user, or you can prompt the caller to enter the agents extension number.   We prefer asking LEX for the user names as it eliminates publishing a directory of extension numbers.  In either case, lambda might convert name to extension number, look up the contact flow details in the database, indexed by the extension number and  “set contact attributes” step with the details needed to populate the rest of the contact flow.

Is Billy Bob available?

Once we have the individual name and extension number of the user we are looking for, we can use the details returned by the lambda function “getQueueDetails” from the DynamoDB table “QueueDetails” to populate and set “check staffing” step.  If the user is available we can send the caller to that user.  If they are not available, we can offer options like “try someone else” or “leave a voice message”.  Voice messages can be sent as wav files, transcribed as text and even sent as SMS messages!    For users that might actually be field sales or support personnel, we can pull back their “cell phone number” from the same database table on the original lambda invocation and pass the call to an external number as yet another option.

Configure Queues as Users?

Using the same resources described above, we can take voice messages for customer service queues.   The “manage users” directory in the contact center administrator portal contains the first name, last name, login and email address of each users.  By adding a queue name as a user in your configuration you can add an email address for that customer service queues.  This means that you can now create options in your customer queue flow that not only offer a call back when an agent becomes available, but you can also encourage them to take a message.  The message converted to email can then be routed to the next available agent using email routing provided by Dextr and agent dashboard for Amazon Connect.  This is also a handy option for after hours call handling!

Post Call Survey with Recorded options!

In our next blog we will use these same resources to enable audio based “post call surveys” that record the callers answers and use transcription, comprehension and sentiment analysis to grade the survey!

If you can imagine it, we can make it happen!

The options for configuring Amazon Connect contact centers is limited only by your imagination.   The AWS ecosystem is populated with an expanding library of services from speech recognition, AI, Bots, text to speech, chat, transcriptions and even language translations.   We believe we are rapidly approaching the possibility of having an English speaking agent interact in near real time with a Spanish speaking caller.

Give us a call and play our favorite game “stump the vendor” and lets hear your call center dream, we think we can make it a reality. –







Configure Amazon Connnect for SSO using Microsoft Azure

Azure AD Configuration

There is work that needs to be done on both sides and typically two different engineers will be working the issue, one on Azure and one in Connect.  Step one is for the Azure engineer to setup a new application using the login link that agents would normally use to login to the Connect instance.  You can find this on the Amazon Connect home page inside the AWS Management console.   The Azure engineer will then provide a Metadata.XML file back to the AWS Engineer.

Step 2 AWS IAM Provider Configuration


  1. Log in to AWS and open the IAM
  2. Click on Identity providers and then Create Provider.
  3. Choose the Provider Type as SAML.
  4. Enter Provider Name, such as “Azure AD”
  5. Upload a Federation Metadata XML (downloaded from previous step).
  6. Click Create Provider


Step 3 AWS IAM Role Configuration ( More Information Here: )

  1. From the IAM/roles console Create a New Role
  2. Select SAML 2.0 Federation trusted entity type
  3. Select the Azure AD SAML provider from previous step
  4. Select Allow Programmatic and AWS Management Console access. The rest will auto-fill.
  5. On the Attach Permissions Policies Page create a policy like this:


“Version”: “2012-10-17”,

“Statement”: [


“Sid”: “Federation”,

“Effect”: “Allow”,

“Action”: “connect:GetFederationToken”,

“Resource”: [






  1. After policy is created, go back to Create Role tab, reload the policy list, and select your new policy.
  2. Set a role name and description, then click Create Role
  3. Open the new role and copy the Role ARN into notepad. Switch to the trust relationships tab and copy the Provider ARN into notepad.

Step 4 Create User for Azure to pull Roles for Users

1 – Create Policy “List Roles”

2 – Create User with programmatic access and attach the policy with the Access and Secret Keys

3 – Send me the Provider ARN and Role ARN back to the Azure engineer along with the User and Access Keys where the balance of the configuration is completed

The Azure engineer will then complete the Provisioning section setting the mode to Auto