2022: The Emergence of the Natural Language Enabled Enterprise

The natural language enabled enterprise is an organization in which humans and artificial intelligence (AI) work side by side to deliver business outcomes with an effectiveness that far surpass what can be achieved when either is deployed independently. It’s a fundamental paradigm shift with immediate consequences for enterprise AI.

Most successful AI projects are based on two presumptions: First, that AI will learn from either an enormous number of labeled data or even greater amounts of unlabeled data; and second is that the resulting trained AI model will replace a human task.

Unfortunately, most unsuccessful AI projects are based on those two assumptions as well, resulting in a 95% failure rate for new AI projects. Failure oftentimes occurs before deployment because human AI specialists have limited influence on models during their creation. They spend most of their time looking for more data and hoping for the best. Failure also happens after deployment because humans are removed from the loop altogether, requiring models to be nearly 100% accurate before they’re useful.

In 2022 and beyond, successful AI projects will involve a cooperative paradigm predicated on two much more dependable presumptions:

1. AI will generate a feedback loop between technology and business savvy humans so both contribute equally to a shared model.
2. AI will produce a cooperative experience in which humans perform existing tasks assisted by AI. Thus, AI models are helpful even if they’re well below 99% accuracy.

Businesses are moving away from the supervised learning approach to a natural language-enabled model that empowers humans to identify opportunities and threats and take more immediate action. The natural language-enabled enterprise provides these benefits via a diverse set of interlocking layers involving state-of-the-art technologies easily employed and tuned by nontechnical users.

The Shared Model Feedback Loop

Human and AI collaborations are fundamental to the natural language-enabled enterprise, and the shared model feedback loop is at its core. This feedback mechanism consists of at least four different layers, the most important of which include:

• Natural Language Understanding (NLU): NLU is a subset of natural language processing (NLP) that focuses on the semantic understanding and underlying intention of words. NLU is instrumental in the first part of the feedback loop in which AI provides value to end users by performing a number of critical functions on text data. The different technologies involved in NLU enable AI to automatically extract information from any corpus by substantially relying on evolving transformer-based techniques. They also perform semantic modeling to understand language variation while driving query resiliency. Finally, NLU has a number of techniques for determining the intent of natural language communication alongside capabilities for explaining results for search, discovery and briefings, so users can tune the system.
• No-code Models: The next layer uses no-code models so humans can input enterprise knowledge into the system as needed. This capability allows people to guide additional information extraction jobs, semantic model evolutions and intent scenarios in partnership with AI.
• Natural Language Services: This third layer relies on natural language services for system diagnosis and zero-code model tuning.
• User Experience: The fourth layer centers on the overall user experience and is based on live testing, benchmarking, curation tools and further diagnosis. It lets humans easily improve the accuracy of AI models via explainability and zero-code tuning. 

A Cooperative Experience: Enabling Humans to Do Their Most Meaningful Work

The natural language enterprise’s cooperative experience between AI and humans is a key benefit of this approach, bettering organizations via cognitive computing. The hallmarks of this experience include delivering immediate value from organizations’ AI investments by enabling them to interact with their text data using natural language to readily mine, explore and profit from it, in layperson’s terms. There are three layers to this aspect of the natural language enabled enterprise. The first involves the delivery of three core natural language services that are integral to the cooperative experience between humans and AI:

• Natural Language Search: Search is one of the most valuable means by which humans can interact with text data. Natural language search technologies are far more effective than simple keywords. This cognitive approach understands the underlying intention of searches and their respective terms to deliver the most meaningful results with question asking and answering. It swiftly and accurately finds information required for knowledge workers to do their jobs.
• Briefing: This service enables people to rapidly skim through a large collection of documents or corpuses based on specific questions of interest to them. Users simply present those questions, which the AI models consider when quickly parsing this information to surface salient points related to them. This functionality lets people learn faster than they otherwise can about a broader array of knowledge or information than they’d otherwise be able to read themselves at any given time.
• Discovery: This service allows end users to find unexpected information in whatever corpus on which they happen to be working. One of the uses of this capability is to inspire new paths of investigation of text data to, for example, determine how an existing body of evidence might affect the outcome of a legal matter.

The second layer of the cooperative experience between humans and AI is about self-diagnostics. This layer lets the system capture and analyze the behavior of the people using it to either correct the system by itself or by working with the humans who influenced the prior model feedback loop. Either way, continual improvement occurs. Lastly, the third layer is based on a user experience that ties together the individual services for a comprehensive cooperative experience. This layer also reports human behavior back to the self-diagnostic system for additional performance improvement opportunities.

In the coming year, and by working with natural language-enabled processes, humans will get the most from their AI and AI gets the most from the humans that are deploying it. The increased productivity, efficiency and effectiveness on both parts enable people to concentrate on their most meaningful work by quickly and effectively applying that information to generate insights and will be a true competitive differentiator in 2022 and beyond.