Ada releases new automated generative AI-driven customer service suite

 Introduction:

The release of the first generation of Ada’s automated chatbot solution and conversational analytics features marks another step towards building an out-of-the-box AI that can handle more complex tasks than its predecessor. At this stage it’s still not ready for production — it still requires human programming, training, and monitoring — but even at this stage, the technology is already saving companies millions of dollars per year. To give some context, Ada will initially charge $1 million upfront to scale engineering costs while also paying developers to add additional capabilities like natural language processing (NLP). But if customers want to buy into Ada’s platform they have to commit to having their own NLP team to make sure every piece of data that Ada collects and makes available to other AI tools sits with a quality product so any data from Ada can now be used for anything under the sun.

Ada is looking forward to continuing its work to take what’s come before it and build on top of its existing products to ensure they’re usable to organizations where these kinds of services are needed. It plans to begin making the platform available in Q3 this year and beyond (adam.ai).

Ada isn’t trying to reinvent the wheel. In 2018, the company partnered with AWS and developed Adex on Amazon Web Services (AWS). As part of its collaboration with AWS, Ada leveraged machine learning to create better conversation transcripts for Alexa Skills. It's the latest offering, the auto-generating AI tool that helps power that functionality is called Ada Conversation Analysis Platform, aka ACE.

The software includes four major components:

• A Conversational Analytics Engine that handles all aspects of sentiment analysis and provides real-time feedback to users regarding the accuracy of AI responses; and
• that handles all aspects of sentiment analysis and provides real-time feedback to users regarding the accuracy of AI responses; and An Engagement Metrics dashboard that shows how well conversations are performing across channels in areas like chatbots and social media. With such visualizations, users can track engagement and improve both performance and results as needed.
• Ada does all this without actually using large language models – those technologies were designed only for very specific applications. Instead, it utilizes pre-trained text-based bots with automatic dialog management and speech processing that are trained based on user input and dialogue data.
• Since people use voice much less than text and understand what’s being said far better than any number of statistics, Ada believes artificial intelligence needs to adapt better to the way humans communicate.
• A good model won’t necessarily translate well, nor is one optimized for various platforms, but rather for a particular type of language. By default, these models run on Amazon Lex or Nuance Communications LUIS, whereas Ada runs its own models on GPT-3A and GPT-3B for Google Voice.

Result:

When an individual speaks back after recording a conversational interaction with Ada through both mobile and website interfaces and then sends it over to the Ada engine to analyze, the system doesn’t just extract relevant information but also contextualizes it to form conclusions about its speaker. For example, the algorithm might recognize that the intent is “I am shopping online and need advice on buying clothes online” and convert that sentiment into something like “You should try [brand] X because it's selling there”. Based on all the knowledge it has gained from analyzing conversations it’ll create its recommendations on Amazon or whatever the source may be. This kind of personalized recommendation system could save businesses hundreds – or even thousands – of dollars annually by not wasting time sifting through calls, texts, emails, and web pages. And the best part? Using AI to help interpret complex statements, it’s able to learn from user inputs and create its own responses based on the patterns found within conversations. That means that although Ada uses sophisticated techniques and algorithms to determine which words mean what or identify a given emotion, it generates its language and intents on its own, keeping it closer to ‘human’ conversation interactions. Plus, not only is it incredibly effective in converting different languages from one another, but it’s also incredibly adaptable. While Ada’s current version is limited to English (and Spanish has yet to join!), it easily detects accents and regional differences as it learns from conversations. All this is done by feeding it human-readable sentences (but not entire paragraphs), and that’s all that Ada cares about. Once it’s finished creating its own response, it shares them with anyone who asks to hear them.

Capability:

When you combine this capability with others to provide valuable insights into customer behavior including sales, marketing, or product development, you end up with intelligent and accurate ways to manage conversations and convert them into action; when combined with predictive maintenance solutions using Ada analytics to understand which leads are the most likely to churn or convert, it saves money; and when combined with smart scheduling and task planning, business leaders can gain valuable insight into employee productivity; and when combined with contact center management systems, many of the most costly issues can be eliminated entirely.

The key here is not getting bogged down in big science or complicated jargon. What matters most for users is that the answers they get are correct, and therefore easy to understand, so they get things done at the speed that matters most. There’s no question that AI today doesn’t look like super advanced deep learning systems in movies (those aren’t nearly as common as we think); but despite that fact, their usefulness is growing exponentially as they become mainstream products. We’re just starting to see the full effect of self-learning and understanding and adapting to conversations as people speak. I expect that progress will continue to accelerate as consumers start talking more with chatbots as opposed to calling it quits, leaving people frustrated. That being said, since the beginning of our journey, Ada hasn’t looked or felt better than in previous iterations. That’s why we feel confident Ada is an important step towards democratizing AI and opening doors for everyone (with and without experience or resources).

Summary:

Ada Conversation Analysis Platform is an automated chatbot solution and conversational analytics features. It's still not ready for production — it still requires human programming, training, and monitoring — but even at this stage, the technology is already saving companies millions of dollars per year. The company plans to begin making the platform available in Q3 this year and beyond (adam.ai).