The Reality of AI — Once You Get Past the Hype
By Donna Fluss
Artificial intelligence (AI) is here, or so we are told every day. The problem is that just yesterday (or a couple of weeks or months ago) most vendors didn’t have AI, and today they say they do. The AI claims might be credible if they were being made by only a few vendors, but when an entire industry of providers is suddenly heralding the arrival of this new technology, it becomes harder to believe.
The first time I heard about AI being included in an application was in 1986, when knowledge management (KM) was introduced to the market. For KM to work back in the 1980s, all of the intelligence had to be built into the application up front. Because the KM solutions required predefined rules, most AI scientists did not consider them to be true examples of AI. While it’s a different story today with a few of the advanced KM solutions, not all of them, despite vendor claims, are actually using real AI.
This brings us to a fundamental issue with the current offerings: Just because software is smart doesn’t mean it is using AI. The debate continues today about whether using pre-established rules disqualifies an application from being called AI-enabled. Software developers say no, but AI scientists typically say yes.
DMG Consulting defines AI as a “branch of computer science focused on the development of computer programs and machines to solve complex problems by simulating behaviors associated with intelligent beings. Specific capabilities include learning, reasoning, problem solving, and self-correction.” At first glance, this definition, like many others, is very broad, which contributes to the misunderstanding and hype surrounding AI. In terms of practical technology, most of the AI that is being used in the real world is designed to solve a well-defined and specific class of issues. Real applications of AI in the customer service/contact center world include the following:
Intelligent virtual agents (IVAs). These are omnichannel solutions that combine natural language understanding (NLU) along with natural language generation and natural language processing to provide a “concierge” level of service to customers. IVAs are being used to replace traditional speech-enabled interactive voice response solutions, as well as to respond to chat conversations and emails.
Analytics-enabled quality management (AQM). These packaged applications are built on a speech/text analytics platform. AQM solutions are able to review 100 percent of interactions received by a contact center and fully automate the process of completing agent quality evaluations without human intervention. These solutions use machine learning to identify new trends as they arise.
Contact center workforce management (WFM). This new generation of solutions applies many algorithms and uses pattern detection to model and optimize forecasts; machine learning is used to identify the algorithm that is the best fit for each round.
LEADING AI PROVIDERS
The best known providers of AI for customer service/contact centers are Amazon, Apple, Google, IBM (Watson), and Microsoft. These vendors use their own solutions and also make them available for third parties to incorporate into their own solutions. There are also more than 75 small and emerging vendors who claim to offer AI technology. DMG has analyzed many of these solutions, and a few appear to provide AI capabilities but do not have traction in the market. Some of them have already disappeared due to lack of revenue, funding, or acquisition. It’s one thing to have a good idea, and another to be able to build and commercialize a viable product.
AI technology is real, but it’s in its infancy when it comes to practical applications. Most vendor claims are hype, which is causing a great deal of confusion in the market. There are large and small providers of AI (e.g., machine learning and NLU), and all of these solutions are highly differentiated. Within 10 years, AI functionality will be a standard component of many applications, but currently, buyers need to carefully assess the claims and the solutions so they can separate hype from reality.