Digital Transformation or Natural Language-Enabled Enterprise

Picture the scene. It’s 2025, and you’re sitting in a digital transformation meeting. There’s pizza on the table, and the coffee is flowing. The person heading the discussion is talking about installing chatbots and transferring paper documents into a digital medium.

‍

There’s something deeply wrong with this picture, and it’s not the pizza (even if it has pineapple on it).

‍

By 2023, the number of large enterprises buying chatbot-only platforms will decrease from 90% today to 10%. It’s hard to believe such a dramatic shift will take place in just a few short years, but the experts are confident it’s going to happen. Natural language technologies are evolving rapidly, and the enterprises that thrive in the future will be natural language-enabled enterprises.

Why Digital Transformation is Out

Digital transformation is defined as “the adoption of digital technology to transform services or businesses, through replacing non-digital or manual processes with digital processes or replacing older digital technology with newer digital technology.”

If we go by this broad definition of digital transformation, then it’s likely to exist for some time. However, in practice, digital transformation usually means something different. The way we define something is typically connected to how we see it operate in practice. The way the companies around us do digital transformation informs what we think digital transformation is.

So what does digital transformation look like in practice? It means tackling business processes one by one in a fragmented fashion. It means relying on relational database schemas across various vendors. It means depending on siloed data sets. This is what is out in 2021. Today’s top enterprises are moving towards a more comprehensive AI model that puts extreme interconnectivity and the latest technology at the forefront.

In the 1500s, Copernicus broke the current understanding of how the world functions and our place in the universe. At the time, people believed in a human-centric view of the universe. They thought that the planets, Sun, and stars revolved around the Earth – the most critical object there ever was; a living, breathing habitat chosen by God. Many people were highly reluctant to let go of this concept and accept that Earth was just one of 9 (at the time – goodbye Pluto) planets that revolved around an unremarkable star we call the Sun.

But why? Because admitting that the Earth (and us by association) was not the center of the universe was admitting we might not be as special as we once thought.

But, what’s this got to do with AI?

Today, we need to move towards a way of working that is less fragmented and less business process-centric. Many people in the 1500s failed to see that there’s beauty in the new model. In an Earth-centric view of the universe, we were special, but it wasn’t earned. We were granted a front seat to trillions of stars. In the Copernicus model, we were special despite our position, not because of it – we thrived against the odds in a brutal universe. Earth earned its place in the universe by utilizing the abundance of resources available to it.

Right now, there is a Copernican shift underway in how enterprises handle data. Enterprises are moving away from an approach that puts relational schemas at the center of AI and towards object models expressed across semantic and machine learning technologies. A natural language future is just around the corner, and enterprises that want to thrive in this future need to take the next step.

Taking the Next Step

In this section, we will discuss the priorities for businesses looking to move towards an NLU-centric enterprise model. By 2024, it’s predicted that up to 80% of branded digital experiences will be delivered to consumers via virtual people. Virtual people are fully actualized conversational AI solutions driven by a wealth of information that is managed intelligently.

Leveraging a Wider Digital Landscape

Traditionally, NLU has been established on a per-business-unit level, but this is about to change. With the rapid acceleration in NLU in regards to chat, document, and email, businesses have been forced to reconsider how they handle their data. Now, the focus is on engineering information so that it may be taken advantage of by machines and humans alike. In 2021, it’s become abundantly clear that today’s publish-centric approach to knowledge management is not appropriate for building language models for modern thriving enterprises.

For most enterprises, speech and language ambitions for consumer interactions are the current primary focus of NLU. However, the tools that support language-enabled enterprises go far beyond speech analytics and chatbots. These tools can transform other areas of the business like customer behavior analytics, process automation, code realization and testing, and much more.

Application leaders wanting to assemble an effective AI ecosystem need to leverage a wider digital landscape, but how? By constructing an enterprise metadata and semantic platform using a composite AI approach. In other words, we need to evolve from a fragmented concept of modeling to a more connected one. There are three ways to represent a concept or object in AI:

Structural – Schemas, ontology, graph, or taxonomy.
Computationally – Utilizing statistical machine learning (ML) or a neural network (NN).
Hybrid/composite – Using a mixture of the two. For example, combining knowledge graphs with neural net classifiers or knowledge graphs.

Leading conversational AI will use a combination of structural and computational technologies to deliver the best results. For example, it could utilize datasets on the ontology of emotions, like physiological responses, character traits, label sequences, and so on. It can also use social context. On the computational side, it can use facial expression classifiers (NN), vocal characteristic classifiers (NN), predictive emotional algorithms (ML), text analytics emotional classifiers, and so on.

Establishing a “Human in the Loop” Program

One of the main challenges of ushering in an AI-centric workforce is employee reluctance to new technologies. Some people believe that getting behind AI is like turkeys voting for Christmas or slugs voting for salt. Just like how a turkey wouldn’t want to support a system that will put it on a dinner plate, employees worry that by supporting AI, they are voting themselves out of a job. However, this is incorrect. A well-functioning AI-centric enterprise has humans at the forefront.

Currently, there is heavy effort involved in developing training data and applications outside of common use cases, but this is predicted to change. Rich human-in-the-loop ecosystems are expected to take AI to the next level by developing and refining a language-based enterprise. Human-in-the-loop approaches are already very popular in the translation industry where people can add invaluable context and nuance to AI-based language translations.

The Gap and Overcoming The Challenges

We’re currently in a state where many enterprises are in a technical and cognitive deficit because language projects are not reused. Because there has been such a heavy focus on per-use-case language projects, business and language data is not consolidated across the business, and data silos remain a prominent issue.

Currently, NLU systems are being built to serve one purpose and not built with an enterprise-wide approach in mind. The architecture across data, design, and underlying AI services are too tightly coupled with not enough room to stretch and expand. This creates a situation where the AI algorithms are too narrow because they were designed for a specific use, and they, therefore, can’t be reused. And lastly, there’s a distinct lack of a mature data pipeline for natural language.

‍

Next Generation Information Architecture Tools

Currently, most informational architecture tools are piecemeal. However, they are starting to evolve to utilize next-generation approaches like data fabrics, semantic platforms powered by ML, and second-generation metadata management. As information architecture tools continue to advance, enterprises will be empowered to create loosely coupled language architecture that eliminates the inflexibility in the current way of working and enables the reusability of language projects.

‍

Roadmap for the Natural Language-Enabled Enterprise

Short Term Priorities

Begin to work on your enterprise graph. If you want to experience powerful AI, you need to move from knowledge management to knowledge engineering. The goal is to build knowledge graphs that enable AI-driven enterprise applications.
Use virtual assistant networks to simplify user interfaces and cover more use cases.
Establish a “Human in the loop” approach. Find human ‘champions’ for your AI solutions to help train, curate, and quality check your NL capabilities.
Create a roadmap for your AI strategy.

Medium Term Priorities

Utilize and combine insight engines (search/discovery) with conversational AI systems. Stive for insight engines with vital integrated touchpoints with conversational AI systems because conversation AI is the future.
Use language analytics systems to determine the health of agent conversations, smart routing, and the overall customer experience.
Stabilize essential definitions and concepts with semantic platforms. Taxonomies, ontological models, and graphs provide a necessary foundational layer for well-functioning AI. Important industry definitions and concepts (education/health etc., terminology) must be mastered centrally and served by bidirectional APIs.

Long Term Priorities

Blend first-party data with third-party open data to supercharge your natural language technologies.
Experiment with “virtual beings”. Play around with your conversational AI’s tone, character, and avatar to develop your future brand character. Virtual beings can be adapted to meet specific customer wants and needs and enhance customer experiences through hyper-personalization.

Where Are We Heading?

The natural language landscape is evolving so rapidly the NLT approach that enterprises have been using over the last few years will look ancient by 2025. To prepare for a natural language future, you have to put the steps in motion today. But where will this lead us?

Systematic intelligence

At its core, a natural language-enabled enterprise is a Copernican shift in handling language. It’s a shift that will transform the customer experience, systematic intelligence, organizational scale, and flexibility. Unifying and reusing your language projects and intelligently handling your data will lead to increasingly powerful conversational AI systems. When it comes to AI, progression isn’t linear. You’ll find that the more you optimize your data for AI, the faster you advance, and the more powerful your AI becomes. It isn’t a one-for-one improvement; small steps forward can create giant leaps in your AI capabilities.

A natural language enabled future is one where:‍

AI understands what was meant rather than what was said. ‍

Multimodal solutions replace solutions-per-mode. ‍

Composite AI is the norm.

AI that is proactive rather than reactive.

'No emotion' AI is replaced with enterprise digital humans.

Most language programs are reusable rather than custom-made.

Recent Innovations in NLU

Engaging digital experiences are a massive driver for NL technology, and this is a great thing. We continually see NL advancing to meet the needs of an increasingly demanding and evolving consumer base that strives for great digital experiences. This has to lead to some notable innovations in NLT:

The advancement of generalized language models like GPT2/3 and BERT that can be used as a basis for creating custom models for industries.
Conversational middleware. These platforms allow organizations to use a flexible combination of speech and conversational engines to support richer conversational experiences.
Multimodal approach – NL firms are increasingly offering multimodal language models that include translation and computer vision to allow for rich and flexible exchanges between computers and humans.
A dedicated shift towards reusable data services, pipelines, and APIs aimed at supporting the next generation of AI.
A shift towards low-code and no-code enterprise tools that put businesses in control and allow for a more flexible approach to managing data and developing applications.
An increased focus on dedicated NLT pipelines and information architecture tools paramount to the future of an NL-enabled enterprise.

How Do You Become a Natural Language Enabled Enterprise

If you found this information in this post exciting but somewhat overwhelming, don’t worry. Becoming a natural language-enabled enterprise is achievable for every forward-thinking enterprise, as long as you have the correct tools and the right people behind you. You don’t have to go on this journey alone. Get in touch with us today, we’d love to help you think through your digital transformation differently.

About the author

Israel Krush CEO & Co-Founder, Hyro

Israel is Hyro’s CEO & Co-Founder. Starting as a software engineer at Intel, he steadily progressed to leading engineering and product teams at various high-profile startups, including Zeekit, a computer vision company acquired by Walmart. Israel’s biggest love (following his wife and three children) is excellent coffee, which serves as the jet fuel for his bigger-than-life ambitions.

Share this post