New Omnichannel Offerings Share the Power of NLU
Organizations are transitioning to omnichannel service, driven by the need to improve customer experience and transaction success rates. Additional business (and IT) drivers are to reduce time to market while cutting tuning and maintenance costs across all channels. And increasingly, that means relying on applications powered by natural language understanding (NLU).
Let’s briefly outline natural language processing (NLP), NLU, and natural language generation (NLG) to ensure a common ground for these terms. NLP processes and analyzes large amounts of audio/voice and text data so computers can respond to language-based commands. My column in the winter 2021 issue—“Speech Technology as Hyperpolyglot?”—delved into machine interpretation of speech. Grammar irregularities, homonyms, and other complications handled by the human brain must be processed by NLU, as it assigns structure, rules, and logic to language so computers can understand what humans say. And NLG conversely generates a humanlike output in text or audio based on structured data. NLU and NLG are subsets of NLP.
Effective omnichannel offerings allow customers to move between channels seamlessly. For example, while a customer might start out with chat, the interaction could move into more complicated topics that the voice channel supports more easily. A good customer experience would include the prior chat conversation so a different agent handling the voice transaction would not have to ask the customer to repeat the details. Some omnichannel vendors are working on providing a sentiment suggestion (angry, excited, etc.) that accompanies a short summary of the previous chat so voice channel agents can be better prepared for the call.
Increasingly, enterprises are moving toward putting more eggs in one basket through single-vendor omnichannel strategies, even organizations with a history of sourcing the best of breed. What is driving this change in IT strategy? Quite simply, effectiveness, defined by businesses as a combination of customer experience and transaction success rates. When customers have chat transactions on your website and are successfully assisted by an intelligent virtual assistant, no human agent required, that’s viewed as a transaction success. Customer feedback about the process drives the customer experience metric.
Many large enterprises have relied on robust interactive voice response (IVR) applications fueled by natural language understanding, with many dependent on their speech recognition vendors to tune their NLU models. But with the addition of artificial intelligence driven by machine learning, constant incremental improvements to their NLU models without direct vendor involvement have caused enterprises to want to repeat these successes in their digital channels. Conversely, some want to extend their NLU learnings from their newer chatbots to legacy IVRs. At this point, it should be noted that deep neural networks are coming of age to supercharge NLU deployed in the cloud; Google, IBM, and Microsoft all offer deep neural networks and NLU in the cloud (Microsoft via its acquisition, Nuance Communications.)
Having a common toolset for managing NLU models across channels decreases training costs and time to market. Extending an existing NLU model from one channel to another, natively sharing it within the same vendor’s toolset, is much more effective than exporting the model (even if possible) and attempting to use it in a different vendor’s solution. Hence the shift from best-of-breed to single-vendor solutions for larger enterprises.
These omnichannel NLU-powered solutions are just beginning to mature; they may be off training wheels but are certainly not out of their teens yet. They are competing on platform integration partnerships rather than as stand-alone offerings, focusing on industry verticals and, soon, their ability to use deep neural networks. Microsoft acquired Nuance for $19.7 billion with the stated intention of leveraging Nuance’s significant healthcare experience. Google has announced that it is working with many of its customers to create a healthcare vertical focus. Oracle completed its acquisition of healthcare IT services and software giant Cerner for $28.3 billion in June, the same month that IBM completed the spinoff of its Watson Health unit to Francisco Partners. Three large players running toward the significant healthcare vertical while one divests should make for an interesting study in the future.
It’s an exciting time for those in customer care. My suggestion is to watch for who jumps out the fastest and furthest with integrations of NLU with industry verticals, powered by deep neural networks. And as this column often reminds, regulations around data privacy and security will come into play, and usually following the technology rather than preparing for it.
Kevin Brown is a customer experience architect at Cognizant with over 25 years of experience designing and delivering speech-enabled solutions. He can be reached at firstname.lastname@example.org.