Training Amelia, Our Cognitive AI System

4 minute read

Ever wonder what it takes to turn an AI system into a full-fledged digital colleague? In this post we walk you through the training process that turns Amelia, our cognitive AI system, into an empathetic task master. Read this article for more information.

Unlike many other software implementations, Artificial Intelligence (AI) solutions require a significant amount of training before they provide business value. AI systems must integrate with all front- and back-end data, and also master processes, specific terminology, and the measured reactions necessary to show empathy toward potentially disgruntled customers.

In this post, we’ll walk companies through the ways in which we train IPsoft’s Amelia, our digital colleague, to provide the most comprehensive and customer-friendly AI platform on the market. This process could be as short as several weeks, or as long as three months, depending on what a business wants to accomplish with AI. We work side-by-side with our clients to train Amelia on initial use cases. Companies are welcome to leverage our expertise for secondary and tertiary use cases, but by that point, they will understand the training process well enough to activate additional use cases without our help. If companies run into any issues during or after training, IPsoft’s professional services are ongoing, and we’ll continue to work with clients at their request to ensure optimal production.

Amelia’s Initial Training

The first step of Amelia’s journey is similar to Neo’s initial martial arts training scene in “The Matrix.” IPsoft uploads an assortment of documents designed to enrich Amelia with new knowledge. This could be as simple as a company’s specific business vocabulary or employee handbook, or it could be as complex as all of a company’s historical customer support data. This process might take new employees years to master, but because Amelia absorbs information at machine speed, she’ll be ready to recognize these terms and phrases within days.

Next, Amelia learns a company’s industry-specific semantic network, and any logic framework she’ll follow in her decision-making. She’ll master a company’s pre-trained classifiers and analytic modules that will help her to conduct business according to industry requirements and regulations. This step of the process is what turns Amelia from a generalist into an industry expert who can maintain company compliance with applicable laws. Although Amelia can be utilized in any vertical industry, she has mastered processes in financial services, insurance, telecommunications, gaming and healthcare.

From there, teams will auto-annotate Amelia’s conversational framework so she can understand utterances and common phrases — essentially, the teams help Amelia speak as humans do. This allows her to complete user interactions, regardless of the topics she might discuss or how users respond. During this process, Amelia will actually advise the teams on the best technique to help her learn and understand user conversations. Our clients also work with IPsoft to teach Amelia to follow standard operating procedures, so she becomes familiar with how a typical business operates.

Amelia is programmed to follow procedure with precision. She will never deviate from the exact processes established for her.

Important side note: If you’re worried about Amelia going rogue and selling off business data to your competitor, the previous step is what guarantees this doesn’t happen. Amelia is programmed to follow procedure with precision. She will never deviate from the exact processes established for her. She can for example approve and reject mortgage applications, or suggest insurance policies; but she’s only able to do this by processing data and making decisions based on the strict parameters for which she’s been programmed.

If Amelia encounters a request that isn’t covered by her training she doesn’t rebel like HAL 9000 from “2001: A Space Odyssey.” Instead, she escalates the request to a human colleague. When Amelia escalates a conversation, she stays connected with her co-worker to determine how the issue was resolved. With your business’ permission, she will then apply that knowledge to any similar conversations and avoid unnecessarily escalating future issues.

Amelia’s Secondary Training

Once Amelia is up and running, our clients work side-by-side with IPsoft’s Conversational Experience Designers. These IPsoft employees are charged with helping create natural dialogue-based conversations. So, if our client is an insurance agency, our client will need to dedicate a team of employees who can come up with a range of dialogue possibilities, including tangents people might take during conversations, idioms people might use when speaking with digital colleagues, and most importantly specific terminology the client will want Amelia to follow when interacting with customers. This is the time for clients to tweak and refine Amelia’s in-conversation skills to ensure that she speaks like an insurance agent, or a credit card claims processor, or a medical professional.

This team will also work with our Cognitive Implementation Engineers to ensure that Amelia has the appropriate emotional response to things customers say or do. What triggers might anger customers? Is Amelia dealing with acceptances and rejections? Will she be handling fraud-related issues? Clients assign a team of employees to work with our engineers to test Amelia’s responses to the many emotional landmines that may come her way when interacting with customers.

Once clients have established a comprehensive list of triggers, they can work with IPsoft to provide appropriate emotional responses. Should Amelia remain firm when telling a customer that he’s been rejected for a policy? Should she take a more understanding and empathetic approach and offer alternative plans? Should she escalate angered customers to human agents? This team goes through every possible scenario to help determine how Amelia should use emotional intelligence to deliver the best and most human service experience.

Once the experience is built and ready for customer interactions, our clients will need to assemble a team of employee super users to interact and engage with the technology and optimize its performance. This team will be focused on working out errors and unnatural language associated with early AI implementations. They’ll see to it that Amelia is answering questions correctly, with emotional intelligence, and following the paths designed for her. After initial success with a small user group, our clients can expand to a larger group of employees. If Amelia passes this final test, she’s ready to start working directly with customers.

Although this is an intensive and comprehensive training process, Amelia can be ready to handle basic tasks in just weeks, and training for a single use case will only take three months, if not less. Amelia’s initial training is typically the most time-intensive part of her deployment. When adding use cases to Amelia’s purview after the initial one, there’s no need to start from scratch, especially if the use cases are within the same vertical industry. The quality of Amelia’s training is what allows her to ultimately deliver positive business outcomes, which is why proper Amelia training is such a critical success factor.


Previous Next

The Intelligent Contact Center

Companies have spent decades implementing Interactive Voice Response (IVR) systems in their call and customer care centers, but they've proven unable to keep up with customers' expectations.

In this white paper, we examine the benefits of an Intelligent Contact Center, where companies utilize Conversational AI-powered virtual agents to provide first-line resolution and support for customers, and augment human employees through AI and automation.

Download our paper to learn the benefits of this approach and why current IVR systems simply will not cut it in today’s hyper-paced digital landscape.

Learn More