Introducing GPT powered Contact Center AI
Introduction: AI in Contact Center
As chatGPT continues to grab attention, businesses are increasingly looking to adopt it into their customer experience strategies. When it comes to contact center applications, AI is increasingly becoming a common tool. AI-powered chatbots are providing an efficient and cost-effective way to offer real-time support to customers. By automating simple tasks and providing instant access to information, bots can help reduce the workload of customer service agents and improve the overall customer experience.
In addition to providing basic customer support, contact center AI tools like bots can also be used for marketing purposes. They can be used to collect data about customer preferences and behavior, which can then be used to personalize marketing messages and improve the effectiveness of marketing campaigns.
The use of contact center AI is growing rapidly, and businesses that fail to adapt may find themselves at a disadvantage. Those that are able to successfully implement contact center align their customer service strategies will be well-positioned to take advantage of the many benefits that it has to offer.
With that said, bots aren't the whole equation.There are types of tasks in contact centers that are currently being done byAI. We, at Commerce.AI, are seeing some common use cases of AI in contact centers. Some are independent and some are intertwined with each other.
● Automatic speech recognition (ASR)
● Agent Assist
● Voice/Text Virtual Agents
● Voice Biometry
● Voice-to-text and text-to-voice transcription
● AI analytics
● Call/Sentiment analysis
● PII/PHI redaction
● Emotional intelligence
● Predictive behavior routing
● Robotic process automation (RPA)
● Predictive Behavioral Routing
● Predictive Analytics
● Real-time guidance
● Knowledge Management
● Workforce automation and training
As we explore the role of AI in contact centers, we need to step back and understand a few basics.
What is AI?
First, what is AI? AI stands for artificial intelligence.Primary difference between AI and pre-AI technology is that AI is learned, whereas past systems were programmed. A fundamental technique called deep-learning that made it all possible for AI to come into existence.
While AI has been around for the last 5-8years, a new powerful type of AI technology is challenging older AI technologies. It is called GPT-3.
What is GPT and how does it work?
In May 2020, OpenAI, an AI research lab(in partnership with Microsoft), created the latest version of an AI-based Natural Language Processing system named GPT-3 that can mimic human language. This 175-billion parameter deep learning language model was trained on larger text datasets that contain hundreds of billions of words.
GPT-3, at its core, is a deep learning model that can give out a sequence of structured text if an input sequence of text is provided. This machine learning (ML) model is designed for text generation functions such as
● Language translation
● Summarizing text.
What GPT-3 is?
● A sophisticated text predictor and generator - Predictive algorithm that looks at a pattern of language and mimics it.
● Once prompted, can generate text/response based on the initial chunk of information.
● Advanced the state of the art inNLP - Mimics tone and style of examples; can deliver creative writing, compose business memos, generate functioning code - can generate languages in all styles
● Comes close to creating plausible real agent responses
What GPT-3 is not?
● Does not have awareness and understanding of the question. Lacks the ability to reason abstractly. Needs to be properly primed by a human.
● If a prompt is not carefully designed, GPT-3 will give poor-quality answers
● Does not learn from a knowledge base without explicitly trained (at least for now :)
● When faced with concepts, content, or even phrasing that the Internet’s corpus of existing text has not prepared it for, it is at a loss.
● It can struggle to maintain a coherent narrative or deliver a meaningful message over more than a few paragraphs.
● Doesn’t know anything about products, services and overall business.
● Does not reason; doesn’t really know if the answers are right or wrong.
Aging Tech vs GPT-3
Previous language models worked in similar ways. But GPT-3 can do things that previous models could not, like write its own language. And, perhaps more important, you can prime it for specific tasks using just a few examples, as opposed to the thousands of examples and several hours of additional training required by its predecessors. Researchers call this “few-shot learning,” and they believe GPT-3 is the first real example of what could be a powerful phenomenon.
“It exhibits a capability that no one thought possible,” said Ilya Sutskever, OpenAI’s chief scientist and a key figure in the rise of artificial intelligence technologies over the past decade. “Any lay person can take this model and provide these examples in about five minutes and get useful behavior out of it.”
So given the characteristics of GPT-3, it offers a variety of pros and cons when it comes to its usage in contact centers.
GPT-3 Pros and Cons
As contact center providers are looking to evaluate this technology, we would like to highlight some of the pros and cons of the technology.
● Automation. Reduce time and resources for once manual tasks. It is really good at forming language/sentence given and proper prompt.
● This opens up an extended set up possibilities/use cases:
○ AgentAssist / Enablement → real-time, virtual agents.
○ Extends the accuracy in smaller tasks such as:
■ Call/chat summary
■ Information Extraction
■ Contextual - small talk
● GPT-3 is trained on historical basic concepts and data. It is not kept up to date on new facts. Therefore, fact based Q&A can result in outdated responses.
● Potential harm of use: Since it is based on vast Internet data, GPT-3 can mimic bias, harmful language, and add to disinformation.
● Does not have deep world-knowledge.That means it does not really understand laws of physics, math and other common sense. This can lead to borderline danger when wrongfully used.
Irrespective of its cons, the amount of benefits and ease of implementation is resulting in increased adoption amongst AI software vendors in contact center space. That leads to the next important question: how should partners and other value-added businesses view this development? And what would be the nature of “services” in the future?
In the future, world will need new tools and approaches to address the accelerated automation delivered by GPT-3 type of technologies. Following are some of the opportunities for value creation in this space.
● Bias Reduction/Removal - training the right dataset:
○ In the future, bots will be fully automated, and agents will only handle complex interactions and issues.
○ Agents will become more like air traffic controllers. Agent training and performance will get increasingly complex.
○ Build conversation flow -conversation design
○ We would need to prepare balanced data sets in order to remove the bias from bot behavior.
○ We would need to increasingly participate in assessing accuracy / quality of automation.
○ Need human testers to assess automation.
● Data Quality:
○ Contextual Empathy. We would need to understand empathy really well in the context of automation. For example what kind of language to use, how to use empathy in different contexts(ecommerce = discount, healthcare = understand patient issue and so on)
○ Tuning. When automation is more standardized, someone who can use tools to tune and course-correct the contact center operation.
○ Implementing automation: we can participate in generating training prompts (just like bot intents) for GPT-3 based systems.
○ Data labeling and data quality. While all the automation is in place, data quality would become a key aspect of customer satisfaction. We can participate in data quality using their domain knowledge.
● Value-added Business Strategy:
○ Customers can extract intelligence from tools like Commerce.AI and provide business value.
■ E.g., using Commerce.AI to measure bot performance, identify up-sell and cross-sell opportunities, increase first call resolution and identify next best action (NBA)
In short, GPT-3 is fundamentally changing the way AI is done. And we are excited to be at the forefront of this revolution.