Skip to Main Content

Amazon re:Invent is one of many Amazon conferences held throughout the year, however, it’s the place to be to get up and close with AWS Product, Developer and Partner teams. Held in Las Vegas this year, the event featured over 50,000 attendees spread over 7 hotels, with 2 Expo halls, many day-long lab events, a builder fair, certification center, over 3500 breakout sessions and 4 keynotes… it sure was a hectic 5 days!

This year was filled with a mind-boggling slew of product and feature releases, which are increasing in cadence as Amazon reacts to customer demand and provides ways to secure and grow their dominance on the cloud.

With a staggering 46% year on year growth, 51.8% market share and an increasing feature set, AWS is proving hard for the competition to catch up.


The keynote from Andy Jassy is well worth the almost 3 hours of viewing time. While there are no specific Amazon Connect references, what is clear is that Machine Learning (ML) and Artificial Intelligence (AI) are front and center. It's my view that over the next few years we'll start to see a paradigm shift in the accessibility of this technology within the Contact Center and Amazon Connect is central to that.

AWS Machine Learning Services Stack

ML/AI is hot right now and has been for a few years. However, personally, I've struggled with this somewhat. I've heard the horror stories that say the machine will replace us and I've also seen "AI" seemingly being added to the title of some services that don't appear to offer anything more than previously. The reality is somewhere in the middle, and what I love about the AWS services is the ability to get hands-on to test and learn with them.


The AWS Machine Learning Services stack is impressive and you're able to come in at any part of the stack as Amazon is building right across all the layers. This means ML and AI are accessible to those without Machine Learning developers. You can consume the Services at the top layer; however, if you have the skills, capacity and desire you can build your own models using the frameworks and hardware available.

The above slides include the notable releases below :

Amazon Elastic Inference - Add GPU to Ec2 for Inference workloads
Amazon Texttract - Vastly improved OCFR service to Emily extract text and data from virtually any document.
Amazon Personalize - ML Service for personalized user recommendation
Amazon Forecast - ML Service for Forecasting
Amazon Comprehend Medical - Makes it easy to extract relevant medical information from unstructured text
Amazon Translate Custom Terminology - Configure Amazon Translate output to use company- and domain-specific vocabulary
AWS ML Market Place - Discover and procure algorithms and model packages from AWS Marketplace, and quickly deploy them in just a few clicks on Amazon SageMaker.

AI within the Amazon Connect eco-system

Looking over the Machine Learning Services stack with the contact center lens on draws my attention to the top row. Amazon Connect was released with Polly, LEX came shortly after, re:Invent 2017 saw Transcribe and Comprehend being released and this year we have Forecast, Personalize and Texttract.


Amazon Connect was released with Amazon Polly from the outset, Polly is an AI Service from AWS.

Using Polly doesn't feel like much because, as consumers, we've become used to expect this level of quality. But for some businesses this has been out of reach. With no training whatsoever, you're able to create a good customer experience by simply using the built-in Text to Speech action blocks within Amazon Connect.



Automatic Speech Recognition (ASR) and Natural Language Understanding (NLU) provided by Amazon Lex was released in August 2017.

Adding Amazon Lex to Amazon Connect provides a Voice First interaction where you're able to create a conversational experience that will hold a customer's attention for 30 minutes! (As explained by the NHSBSA Team in HLC305)

One of my favourite sessions was BAP322 - Bring the Power of AI to Your Amazon Connect Contact Center. Marc (AWS), Dan and Laura (both Liberty Mutual) laid out their journey of using Lex, describing how the development team adapted, the iterations they went through to refine the Lex Bot with VoiceFoundry's UX Expert Support, and finally the business value. I find it so important to focus on the WHY. Yes, it's cool and vogue to be working on this technology, but it's short-lived without value. Dan did a great job walking through the experience with the transfixed audience.

Amazon Transcribe and Comprehend


Amazon Transcribe takes the audio after the call is completed and turns it into Text. This allows you to run analytics on that text to feed those all-important feedback loops to find out why the customer was calling. Once the conversation is in text format, you can then run it through comprehend to identify key phases and agent/customer sentiment (Negative/Neutral/Positive or Mixed). We've been able to do this for about a year now, however, with the addition of Amazon Comprehend Medical & Amazon Translate Custom Terminology we will start to see higher accuracy on industry/company specific language.

We saw a glimpse of what's to come... imagine the above but in near real-time. The agent is having a conversation with a customer about an order and the order pops up on screen...the AI Assisted agent is coming. And I personally believe this is an area of great potential and one I'm keen to get into.

Amazon Forecast

Amazon Forecast was announced for preview. I haven't seen it yet but it promises to be a "fully managed service that uses machine learning to deliver highly accurate forecasts," useful for "planning for the right level of available resources, such as staffing levels. Once you provide your data, Amazon Forecast will automatically examine it, identify what is meaningful, and produce a forecasting model capable of making predictions that are up to 50% more accurate than looking at time series data alone."

We'll have to come back to this one at a later date.

Amazon Personalize

This was also announced for preview. It's based on the same technology as - to provide real-time customer specific recommendations across all customer channels by analyzing data in real-time. Amazon has not indicated how this might be used within the contact center, however, I can think of examples ranging from trivial items like queue music (Boring, I hear you cry!) through personalized offers during a natural conversation with Amazon Connect and Lex.

Imagine the scenario of a customer contacting a company to re-arrange their flight (a popular demo). The flight gets re-arranged and Amazon Personalize knows that the customer normally buys an upgrade, so after checking if it's available the upgrade is offered and the customer is very happy. Win/Win!

Amazon Texttract


Amazon Texttract is "OCR++". Not what you might first think about when considering Amazon Connect. However, letters are still a much-used communication channel and thus should be considered within your customer experience strategy. I'm not aware if the Amazon Connect teams are looking at ways to bring this non-digital channel into the Connect Eco-System but I wouldn't be surprised to see this within a few years. That said, you could build a solution using Texttract, Comprehend and the Connect OB API today. I'd love to hear if you have example use cases for Texttract for the Contact Center.

Amazon Connect Breakout Sessions

There were many breakout sessions offered for Amazon Connect but the sessions had to be booked well in advance - normally on the day the agenda is opened up which was many weeks before re:Invent! However many of the sessions were recorded and all of the slides are provided to the public after the event. Below is a great selection of Amazon Connect related sessions.

AIM304 - Transform the Modern Contact Center Using Machine Learning and Analytics - Slides

AIM314 - Create a "Question and Answer" Bot with Amazon Lex and Amazon Alexa - Slides

BAP309 - Customizing Your Amazon Connect Contact Center - Slides

BAP322 - Bring the Power of AI to Your Amazon Connect Contact Center - Slides

BAP324 - Moving Large Scale Contact Centers to Amazon Connect - Slides

BAP328 - Architectures for Gaining Data Insights into Your Contact Center Experience - Slides

BAP348 - Enterprise Strategies and Best Practices for Migrating Your Contact Center to Amazon - Slides

BAP401 - Build a Voice-Based Chatbot for Your Amazon ConnectContact Center - Slides

FSV301 - Financial Svcs: Build Customer-Centric Contact Centers with AmazonConnect and Machine Learning - Slides

FSV306-S - Implementation of Amazon Connect, Powered by Accenture - Slides

HLC305 - Optimizing Healthcare Call Centers with Natural Language Understanding - Slides

TLC350 - Accelerate Digital Transformation for Telecom Operators with Cloud-Native Amdocs Solutions - Slides

Amazon Customer & Partner Appreciation Event

I have spoken a lot about the Amazon Services, after all, reInvent was an Amazon conference. But on Tuesday evening VoiceFoundry, other Partners and a few hundred customers attended the Amazon Connect Customer Appreciation Event.


We heard from many customers on the value that Amazon Connect has brought them, from the Amazon Connect leadership and viewed another demo on the real-time transcription.

The Amazon Connect product is enjoying a healthy growth rate:

900%+ Growth in usage since re:Invent 2017

And with one of Amazon Connect's key principles, being an open platform, the Partner eco-system is healthy and growing constantly.

A personal highlight was seeing VoiceFoundry named as one of only four consultancy partners to be an approved Reseller! Below was the floor plan for the 3 hour event. Since then, re:Invent, Aspect WFM has announced their support for Amazon Connect.


Other notable releases:

Amazon Outpost - Run AWS/Hybrid within your own datacenter, plug and play!
Amazon Quantum Ledger Database (QLDB) - Actionable blockchain service.

Indulge me

Formula 1 is the only sport that I do my best not to miss, so when Ross Brawn came on stage to talk about how F1 is using HPC and ML to improve the sport, it was a personal highlight.