Using Amazon Connect data to speed amazon lex bot deployments

Extract intents, flows, entities and training data from Amazon Connect call recordings and use them to help you build your Amazon Lex bots

By Jonathan Eisenzopf, CTO & EVP Strategy

Happy Lady.jpeg

What if

you could glean your bot strategy, design insights and training data from your existing Amazon Connect conversations?

If you think about it, every customer conversation you’ll have with your customers today probably happened yesterday and the day before. Why not use those very conversations to build better amazon lex bots faster?

Even when using the best bot building platforms, there is typically a lot of manual effort (and therefore time and money) spent in just about every stage of launching a chatbot. But, what if you could streamline this process by taking advantage of the rich set of customer conversation data you already own to inform the strategy, design and training of your bots?

Amazon 3 Step wConnect.png Cognition + Amazon Lex Make it happen.

Leverage customer conversation data to speed your time to value with Amazon Lex chatbot deployments.

Our customers tell us it can take up to 6 months to design, develop, train and deploy one bot intent, with developing good bot training data being the proverbial “long pole in the tent”. Once that bot is launched, they begin the wait-and-see game to see how successful it is with their customers. And, when it fails? Well, that’s when the tuning really starts, because, after all, we all know that “If it ain’t broke, don’t fix it.”

Yes, most bots are trained on the backs of failed customer experiences... See what falls out of the bot, then train it some more.

Amazon Lex provides the platform to build and deploy chatbots in minutes - but you need to first be prepared with insights and a rich set of bot training data to really take the most advantage of it. And you can't risk launching a bad bot. So training and tuning is critical to the success of the bot and, ultimately, to your customer's experience.

This is the gap that is filling with the integration of its Cognition platform with Amazon Lex, where Amazon Connect customers can:

  • Load your transcribed Amazon Connect voice calls, live chat and even chatbot conversations into the Cognition engine to enjoy automatic labeling of the conversation data. That step alone can be a huge cost savings considering outsourced data labeling can cost $2-$5 for every conversation.

  • Visualize ALL of your conversations to see common paths, customer behaviors, patterns and outcomes revealed.

  • Drill down and around in your conversation data to discover opportunities for automation, and inform bot flows.

  • Filter and export conversations and load hundreds if not thousands of intents and utterances directly into your Amazon Lex bot.

NOW you're armed to build an amazing Amazon Lex bot in minutes…

But, wait – there’s more! Once your bots are launched, why not see what you can learn from them? uses AWS Lambda functions to capture information from Amazon Lex bots in real-time to make bot behaviors and flows immediately visible in the Cognition Conversation Insights dashboard.

Laptop-Conversation-Insights (1).png

Watch how your
Amazon Lex bots are performing…
in near real-time!

Feel the impact of your intelligent automation strategy sooner, rather than later.

We’ve already demonstrated a 50 – 75% time and cost savings for the end-to-end bot deployment process using the Cognition platform to augment the Amazon Lex bot development framework. And the real business impact comes with the acceleration of your time to value with new and smarter chatbots. Chatbots that have learned directly from your customers…

Stop by to see us at the Amazon Booth #2106 at Enterprise Connect in Orlando.

We’re excited to be announcing the integration of Cognition to Amazon Lex at Enterprise Connect in Orlando. Come see us co-present with Amazon at 3:40 on Wednesday, March 20th at Booth #2106 and learn how, with Cognition, you can Launch Amazon Lex bots faster using Amazon Connect data.