Oct 03, 2022
As I watch the contact center industry promote everything and anything AI, I’m surprised at how often AI’s benefits are drastically misrepresented. I see and hear many discussions of the benefits of AI that are arguments for automation – with no mention of specific AI benefits over other forms of automation. “Decrease costs, make customers happy” are claims that seem too good to be true and don’t provide much information on what about AI enables that to happen.
AI-based systems, when appropriately utilized, have tangible benefits, especially when compared to the older speech-recognition-based IVR systems (or, heaven forbid, touch-tone DTMF systems), they are replacing. But I feel we are doing a disservice to AI when we don’t describe how the AI part makes the system better.
First, let me share a little history. IVR (Interactive Voice Response) is a technology that has been around for 40 years. At first, these systems accepted touch-tone (DTMF) input, the ubiquitous “press 1 for sales, 2 for customer service….” call directing systems that replaced human operators. Then, the systems allowed data entry over the phone (“enter your account number…”). Back-end data transactions (screen-scraping mainframe terminal screens at first, then more modern API methods) would pull data to the caller for self-service (“your current balance is …”). The return on investment for these systems was straightforward: $5.00 per minute for a human agent interaction vs. $.25 per minute for an IVR interaction (typical costs in the 1990s). If you could automate a high enough percentage of calls, the IVR systems would pay for themselves in months.
Some industries, like retail banking, could see call automation rates over 90%. The cost avoidance potential made IVR systems must-have tools for call centers. They were great at cost reduction – but terrible for customer service. The complexity of end-user needs pushed against the limits of the technology – trees of menus sprouted everywhere, and the term “IVR Hell” came to be known.
The next major step forward was the advent of speech recognition technology. Instead of requiring callers to press buttons, systems could accept spoken input. Speech recognition technology had its limitations: allowing a wide variety of inputs required speech systems trained to a single voice. Dictation-type systems where you could talk instead of type could use this technology, but not in call centers. Speech recognition technology could recognize speech input from a general population only if a small number of specific words were allowed as input. As technology progressed, the number of different words the system could match against the input increased – but it always had some upper limit.
“Grammar” was the term used to define the permitted words at any speech input state. These “directed dialog” IVR systems would use these grammars when prompting for input, collect that information from the caller, and then go to the next input step. If the caller speaks one of the allowed words, the input is accepted – otherwise, the common “I’m sorry I didn’t get that” is spoken to get the caller to try again.
Allowing speech input was an improvement over DTMF-only IVR systems. System designers could develop natural speech input patterns, especially for “form filling” use cases where multiple data elements were needed to complete a transaction. Caller acceptance was higher than with DTMF-only systems. But, technology acquisition, development, and upkeep costs could be three times higher than a DTMF-only system – and the technical limitations still frustrated callers. These IVR systems still forced callers to figure out how to get through the prompts – to put their problems into words the system designer used. Well-designed systems made that more manageable – but too many systems struggled to provide a good user experience.
So this brings us to where AI became involved. Two distinct industry trends converged, bringing AI-powered systems to the contact center. The first was web chat: pop-up boxes on websites connected users to live human agents. These systems were a hit with both end users and businesses, and web chat quickly overtook email as the primary channel for text-based communication in the Contact Center. Companies quickly recognized the need for automation of these interactions, the all-text communication path was straightforward and cheap to build, and the “dumb chatbot” was born. A desire to create better automated text interactions drove the adoption of AI systems with our first describable advantage of AI systems used in the contact center.
For our AI-driven chatbots, end users could ask their questions the same way they would if they were typing their questions to a human agent. The AI-powered system could pull the meaning out of the conversation without restricting the input to a specific form or grammar. That meaning usually takes the form of the “intent” of the interaction, plus any “entities” contained within the conversation. “What’s the current balance of my checking account” yields an intent of “current balance” and an entity of “checking account.” With that meaning pulled out of the interaction, an automated system has the information it needs to respond correctly. A trained AI system demonstrates its value by being more robust and accurate than hand-coded word-spotting or other rules-based algorithms without restrictions or “guidance” from the system on the input. The user interacts naturally, and the AI system figures it out.
The use of AI is not magic - a system will have a pre-defined set of intents it can react to (you don’t expect a banking system to be able to collect information needed to order a pizza…). But, it allows for a more natural interface for the end user and a more accurate and easily maintainable system for the business.
Now that we have AI-based technology for understanding human conversation happening over a text channel, the 2nd advantage of AI technology for contact centers is applicable.
Anthony Brown - stock.adobe.com
Amazon’s Alexa was the first widely used, commercially successful speech input system that utilized new AI-driven technology to break the boundaries of previous speech recognition systems. Amazon’s Echo devices used a combination of advanced microphone technology with an AI-backed recognition engine to pull actionable intents out of end-user inputs in a way that previous phone-based systems could not. Several vendors quickly made this technology available for phone-based interactions (including Amazon!), and it is now the new standard toolset for applying automation to the contact center.
Like with text, once systems can understand the meaning of a customer inquiry, automation tools provide a means to assist with completing the interaction.
So, in summary, here are the key benefits of AI technology for the Contact Center:
AI tools programmatically (without human intervention) determine the intent or purpose of a customer’s interaction over voice or text using a natural interface that does not hinder the customer.
Contact Centers can use this understanding of the customer to apply automation wherever customer communication is in text or voice. Contact Centers are using this automation successfully at several touchpoints, including:
The AI-based systems that do this work are less expensive to build, implement, and maintain – and are more accurate than their predecessors.
AI excels for use cases where your business needs to figure out what your customer is trying to accomplish when they contact you without alienating them by asking 20 questions. And while it does provide a more economical way to utilize speech input for phone-based systems, it doesn’t measurably help you with simple data capture tasks such as collecting an account number. Recognizing where AI provides leverage to your use cases can help you make more informed vendor selections for tools and services.
Enabling automation is still the core of helping Contact Centers work more efficiently. Using the best tools to build that automation allows systems to perform their automated tasks in the most customer-pleasing way. And systems pleasing to the end user get used more and detract less from the customer experience. Recognize AI for what it is – a tool for building better automation. By itself, it doesn’t do anything.
Need additional help? I can provide services to assist in use case development, vendor selection, and implementation guidance. Please reach out!