Experts predict NLP to be biggest BI trend this year

Advancements in NLP

NLP holds great promise for analytics.

Ideally, a business user could type or speak a query using any phrasing in any language, and their BI platform could respond by providing relevant data in a digestible format such as a chart or graph accompanied by a detailed explanation. In addition, the platform would be capable of answering more detailed follow-up questions in the same way so the business user could drill deeper into their organization's data.

But NLP has not reached that ideal state.

Language -- for example, different ways of phrasing the same question and different words that have the same meaning and words spelled the same that have different meanings -- is still too complex for existing technology that must take natural language, translate it into code to run a query, and then translate a coded response back into natural language for consumption.

In addition, there are more than 5,000 languages worldwide. Even programming a platform to converse in some of the most commonly spoken languages is time consuming and complex.

But technology is getting better. And it could result in significant advancements in the use and capabilities of NLP in 2023.

"Natural language generation," said Ritesh Ramesh, COO of healthcare consulting firm MDAudit and a customer of analytics vendor ThoughtSpot, when asked what will be a major analytics trend in 2023. "The ability to automatically generate business insights and commentary behind the insights [will be a trend]. It's technology that auto-generates storytelling from the visualization."

In particular, the release of ChatGPT -- which builds on the capabilities of GPT-3 that was first launched in 2020 -- could spur widespread improvement of NLP in analytics.

Some vendors have developed their own NLP technology. For example, ThoughtSpot had NLP capabilities when it first launched in 2015, AWS built QuickSight Q with Amazon's machine learning technology, and Qlik added NLP capabilities with the acquisition of CrunchBot in 2019.

Now, others have the potential to add NLP capabilities more easily by integrating ChatGPT or another chatbot tool.

Like GPT-3, ChatGPT translates the typed or spoken word into code, runs a requested query, and translates the response back into natural language. But ChatGPT does so in a way that represents a technological leap rather than an incremental improvement.

"We're seeing huge natural language query and generation innovation -- there's leapfrogging going on," said Dan Sommer, senior director and global market intelligence lead at Qlik. "ChatGPT is extremely powerful, and it shows that this natural language chatbot interface will move the dial in the short term. It changes how we can interact with data and how data is generated for us."

Amazon, Microsoft and Salesforce -- parent company of Tableau -- are expected to make substantial improvements in their NLP capabilities over the next 12 months, according to Farmer.

"These [new chatbots] can generate SQL from a natural language input, making them more advanced than the previous generation of natural language query technology," he said. "This is something that is likely to be at the forefront of development over the coming year."

There is, however, still a long way to go before NLP reaches its ideal state, according to David Menninger, an analyst at Ventana Research.

Like other NLP technology, even ChatGPT doesn't actually understand language. It's still a computer program and can only do what it's programmed to do. That does not include interpreting meaning when a query isn't clearly stated.

In addition, responses aren't always accurate and -- like other AI -- it is subject to bias.

"NLP still has a ways to go before it's mainstream. But I think that [evolution] will continue to happen during 2023," Menninger said. "There are still lots of gaps."

While 2023 isn't the year NLP will completely transform BI, NLP technology will improve over the next 12 months, he continued.

"The things that I think will happen with NLP will be things like more free-flowing questions as opposed to structured questions, and I think multi-lingual will become more and more prevalent during 2023," Menniger said.

In addition, he noted that the NLP advancements he expects this year will center around chatbots -- the written word -- rather than the spoken word.

"I still think voice is going to lag," he said. "It will still be more chatbot oriented, but we'll see more progression on the voice front."

Decision intelligence

While NLP will be a dominant trend in analytics over the next year, it won't be the only one.

One that rose to prominence in 2022 and is expected to continue gaining momentum in 2023 is decision intelligence.

Decision intelligence is the use of augmented analytics and machine learning to aid humans as they make business decisions. With the amount of data organizations collect growing at an exponential rate, there is too much data for even a team of data experts to keep tabs on and notice every change or anomaly.

AI and ML, however, can be trained to monitor key metrics and other data for changes, anomalies or trends; surface those insights; and alert data experts who can subsequently determine if actions need to be taken.

ChatGPT is extremely powerful, and it shows that this natural language chatbot interface will move the dial in the short term. It changes how we can interact with data, and how data is generated for us.

Vendors focusing on decision intelligence include Pyramid Analytics, Sisu Data and Tellius, all of which have attracted venture capital funding within the last 16 months despite worsening conditions in the capital markets.

Pyramid raised $120 million in venture capital funding in May 2022, Tellius raised $16 million in October 2022 and Sisu last raised $62 million in late 2021.

"There are a lot of opportunities to help organizations make more intelligent decisions," Menninger said. "I think we're early in the trend, but … enterprises do not get enough support today from vendors and technology, in general."

That plays into what Krishna Roy, an analyst at 451 Research, terms actionable intelligence, which she said she expects to figure prominently in business anlaytics in 2023.

While not specifically decision intelligence, actionable intelligence is related to decision intelligence in the sense that it's the use of technology to support decisions that lead to actions. But rather than surface insights for data experts, the concept is instead about presenting insights within business users' workflows.

"Actionable intelligence will be a big trend," Roy said. "Data-driven decision-making will need to be met using various approaches, depending on the use case and user. Actionable analytics, as the name suggests, will enable this by allowing individuals to execute more easily on analysis by providing it in … familiar workflows they use regularly."

Qlik is among the vendors making actionable intelligence a priority.

Data mesh

NLP and decision intelligence are both focused on using technology to augment a human's interaction with data.

Another trend gaining momentum has less to do with technology but is instead a philosophical approach to analytics.

Data mesh, first introduced in 2019, is an approach to analytics that decentralizes data. Though its benefits have been known for a few years, adoption has been slow. That could change in 2023 as organizational leaders seek new ways of making analytics a larger part of their enterprise.

Traditionally, organizations house data in centralized repositories overseen by a team of data experts and the data is parsed out as needed. In some cases, even the analysis is done by a centralized team that delivers reports and dashboards upon request.

A data mesh approach instead enables teams within different domains, like finance and marketing, to oversee and analyze their own data while connecting each domain's data through a data catalog.

The idea is to take advantage of the domain knowledge of an organization's employees, operating under the assumption that a data expert in finance or marketing will be more knowledgeable about finance or marketing data than a centralized data analyst.

And like NLP and decision intelligence, data mesh also is aimed at expanding the use of data to inform decision-making within organizations. It encourages the idea that domain experts can more easily work with and teach self-service users how to work with data than a centralized team.

"We're going to actually see that the folks who were talking about decentralization a year, two years -- maybe even three years ago -- were right," said Russell Christopher, director of product strategy at data mesh specialist Starburst.

Christopher, whose experience includes six years at Tableau and 14 years at Microsoft, related that he had to show people the benefits of data visualization at one time before they eventually realized its worth. Now vendors like Starburst, Talend and Denodo -- whose tools enable data mesh -- are doing the same for data decentralization.

"Now the lightbulbs are beginning to turn on," he said.

Farmer noted that he's also seeing organizations explore data mesh but added that it's still in its nascent stages. He cautioned that more data governance needs to be includes in data mesh before it becomes viable on a large scale.

"Generally, data mesh is a type of self-service BI, and a few are exploring the architecture more rigorously," he said. "But a major issue that still needs to be addressed is governance, which will likely hinder enterprises from embracing the concept."