Amazon Connect Lessons Learnt

Amazon Connect Lessons Learnt Blog Post.

Amazon Connect is a baby in comparison to other vendors in the contact centre world. And Route One Connect picked up Amazon Connect right from the start so naturally we have gone on a journey and grown with the technology. As such, we have a bunch of Amazon Connect Lessons learnt that we there is value sharing. For Instance:


Amazon Connect


Nuance Comparison

At Route One Connect we have an extremely rich heritage of using Nuance. As such every step of the way we compared Amazon’s Contact Centre offering with Nuance who are known to be the best speech recognition engine available. However on occasions this has caused incorrect use of Amazon Lex. For Instance, confidence scoring is not a score of the utterance. Instead it is a score in confidence of the intent. Another example would include: NBEST. With Nuance you were provided a list of result where as with Amazon Lex you are provided a list of alternative intents. Again a huge difference.

On the other hand, Amazon uses slot much better than Nuance which enables the speech developer or scientist to deploy much simpler grammars but pack the same punch.

Further more, the tuning process between Nuance and Amazon Lex could not be more different. Having said that, having the amount of experience building Nuance grammars as we do has really set us apart from our competitors. Above all, our speech and data science experience means we can deploy highly performant Amazon Lex bots.


Amazon Connect


This is not a lesson we learnt the hard way. We inherited an Amazon Connect and Amazon Lex solution and the incumbent clearly did not tune the Amazon Lex Bot. You must never attempt build a Amazon Lex bot if there is no budge to tune and tune regularly. This is a huge school boy error! Needless to say, we brought a massive increase to performance which resulted in thousands of calls being handled within Amazon Connect and not by an agent with our customer recognising our improvements had reduced the cost per contact into the contact centre.


Amazon Connect


Treat Your Voice Bot Like You Would A Real Agent

AI is a wonderful feature of Amazon’s Contact Centre suite. And it comes in many forms from the likes of Amazon Wisdom to the most obvious AI features being Amazon Lex. The problem with AI it requires training, growth, check up and much like a real human needs continued investment. The best advice we advise our customers is to treat an Amazon Lex boy like a human. You must look after it and invest as much time in it as possible.




Other Observations – AWS is aggressive in it’s growth plans.

We have to acknowledge the growth Amazon Connect and Amazon Lex has been through. It has felt at times that AWS have been making improvements on a daily or weekly basis. However, with AWS being a true cloud platform and managed by AWS none of the improvements have come at a cost or outage. With legacy platforms a patch, service pack or upgrade all came a huge cost either through out of hours change requests or a financial cost for the upgrade. Sometimes new hardware was even required to facilitate the upgrades. However, new features, patches etc under the AWS platform all managed seamlessly.




Finally, if you liked this blog post, please check out these as well:

Amazon Connect Troubleshooting

Conversational AI with Amazon Lex

Protect Vulnerable Customers With Amazon Connect

Amazon Lex vs Nuance 10 – Which is best?

Lets Connect…The Amazon Way

Amazon Lex Release Note