Could LLMs be the ideal no-code data interface?

Two emerging innovations are converging into one powerful opportunity for data analytics services, if one recent demonstration from SAS offers a glimpse of the future. By combining large language models with workplace chat as a way to query SAS Viya AI analytics platform, deep insights into corporate analytics can become a seamless part of team strategy discussions.

Discussing the future of generative AI with SAS Australia and New Zealand Director of Customer Advisory Ray Greenwood, he described a recent innovation showcase which provided an interactive experience for supporting data-driven decision making for non-technical business users.

“This is not just a one-to-one relationship with ChatGPT. For us, it’s about how we leverage conversational AI,” said Greenwood. “When we reduce the interaction with AI down to conversations, we have eliminated all the barriers to adoption. Human language becomes the interface.”

In the demonstration, a chat service tags ‘@viya’ to send natural language queries to SAS Viya. In response to a question about maintaining 10% year-over-year revenue growth, Viya suggests an analysis of the next three years of point-of-sale data. It points out what it expects it can forecast and that the data lives on Snowflake, and asks the user to agree to execute the query.

“We get Viya to seek permission because responding to these queries will result in resource consumption in the cloud,” Greenwood said. Once agreed, Viya performs the review and also tells the chat that it sees anomalies and outliers, so asks if it should do seasonal adjustment to take these out. It follows up with further proactive suggestions, including whether the team is aware of suppliers who are not meeting demand and how to manage shortfalls.

The demonstration offered a fascinating perspective on how the chat interface may offer a simplification for general business users, but the queries were still interpreted into highly performant analytical investigations by the SAS Viya platform. Greenwood argues that this speaks to some of the most powerful opportunities for AI in the future.

“We believe we’re going to have some of the most meaningful impacts in terms of generative AI by embedding capabilities into our industry solutions,” said Greenwood. “We can go from asking questions in the midst of a team workflow to codifying the question into an AI query, and then to communicate that AI output in a way that feels natural in our day-to-day experience of work. That’s powerful.”

We are in the midst of serious trust and privacy concerns over using public language model tools with corporate data. Seeing an LLM in this new context offers a positive sense of how such tools will not simply exist as standalone services for much longer. Instead, they are likely to become the intelligent interface to private data and simplify the application of thousands of other custom AI models designed to serve specific analytical needs. For companies like SAS, built on decades of data analytics expertise, the rise of the LLM is not a competitor to be feared but a capability to be embraced.

“This missing part of the generative AI story today is the ability to apply those technologies to the customer’s data without it going back out to the public models, and in a way that provides transparency and guardrails for users in support of local guidelines such as the Digital Transformation Agency’s guidance on GenAI which references these issues in the context of privacy, IP and ongoing monitoring of model behaviour,” said Greenwood. “That’s where we’re focused and given our ties into the banking industry and government services, we believe we’ve got such a huge opportunity to make this a real, tangible value proposition for those organisations.”

Image credit: iStock.com/GOCMEN