How GenAI Impacts Augmented Analytics

Analytics and business intelligence platforms continue to evolve in many ways, such as expanding the number of people who can use them, surfacing insights faster, and improving the accuracy and value of results. The label modern analytics and BI (ABI) platforms have been given for the last few years is “augmented analytics” because vendors have added AI to their products so humans could augment their own analyses with the help of AI. 

Before generative AI (GenAI) hit the scene, ABI platforms were already using natural language processing and machine learning to understand queries and explain analytical results. Since then, GenAI capabilities have been added to platforms, though they tend to vary in their level of AI and GenAI maturity.  

In the end, the goal is the same: to provide timely, accurate insights to more users. 

The Change GenAI Enables 

GenAI allows organizations to do more with their ABI platforms, such as generating data stories, metadata and code; creating executive summaries and storyboards; and providing automated AI-powered insights, with or without a dashboard. One can also use GenAI to interrogate data. These advancements are made possible using proprietary and third-party LLMs, such as from OpenAI, Microsoft, and Google. Platform providers also continue to acquire point solutions that make for more comprehensive ABI offerings, which now include GenAI companies. 

Related:Addressing Trending Questions About Generative AI

The bottom line is that organizations can do more with data, not the least of which is continuing a conversation that maintains context. 

“Augmented analytics is one of the primary ways that we use AI. While we stay away from generative AI for content creation or communication, AI can be a fantastic tool when it comes to data analysis,” says Edward Tian, CEO of GPTZero, which detects GenAI in written content.  “Something in particular that stands out is generat[ing] really valuable insights. It can find patterns and use algorithms to identify insights that we may not have ever noticed on our own, and that helps us make the most informed decisions possible.” 

Homegrown Solutions Are Also Benefiting

SMB lending platform Credibly was able to successfully augment its internal analytics using GenAI. The company experimented with vector databases and retrieval augmented generation to develop more robust business profiles of its applications. According to Ryan Rosett, co-CEO and founder, these kinds of enhanced analytics, through combining GenAI and supervised models, have led to shorter approval times, better accuracy, and a deeper understanding of the customer.  

Related:How Data, Analytics & AI Shaped 2020, and Will Impact 2021

“GenAI is taking Credibly’s internal business intelligence efforts to another level,” says Rosett in an email interview.  

The most important innovation for his company has been finding real use cases that couple GenAI and supervised models to produce more accurate results.  

“We didn't jump on the AI bandwagon and start throwing all kinds of use cases at it to see what sticks. We are constantly experimenting, comparing results, and figuring out how GenAI can work with existing models, and how they can make each other better,” says Rosett. 

As a lending platform, Credibly must be able to conduct a fast and accurate risk assessment of business owners seeking financing. To achieve that, the company developed a methodology to risk-adjust external data, created a proprietary search engine using GenAI that quickly ingests and summarizes metadata from external and internal sources, and paired that with automated machine learning models to provide more accurate, risk-adjusted determinations for underwriting use. 

“One of the major benefits has been to increase speed and accuracy, while eliminating the costs associated with manual review so that our underwriters can be more productive. Per employee productivity and revenue increased when we layered in multiple use cases,” says Rosett. “In one example, we were able to reduce complexity from using a couple thousand selections down to fewer than a hundred, and improved search time from a few minutes per deal to less than 30 seconds [by incorporating] automation. In another use case, we were able to reduce our offshore footprint by bringing the task back to onshore employees through removing a manual step in the process and using the GenAI model to deliver recommendations for review.” 

Related:How Safe and Secure Is GenAI Really?

Of course, the journey hasn’t been all roses. GenAI hallucination mitigation, nondeterminism, and user adoption can be challenging.  

“It's important to keep users engaged while deploying GenAI models [because] there is a ton of training involved,” says Rosett. “Using existing databases and supervised models to help rank answers also help reduce the issues associated with [hallucination mitigation and nondeterminism].”  

Bottom Line 

GenAI is making its way into all types of applications, including analytics and BI. Its addition enables greater natural language capabilities that benefit data scientists, analysts and citizen data scientists. 

The GenAI capabilities and to the extent they’re used varies from vendor to vendor and within enterprises, but things are moving swiftly. GenAI may be a temporary competitive differentiator, but it will soon become a commodity in the ABI space like predictive analytics, data visualizations, and dashboards.