Register for "A Guide to Data Products: Everything You Need to Understand, Plan, and Implement" - Friday, May 31,1:00 p.m. ET

Driving Results with Conversational BI: Best Practices for Power and Casual Users

ABSTRACT: This blog defines conversational BI, why companies should consider it, and how their power and casual users can best get the desired results.

Read time: 5 mins.

Sponsored by ThoughtSpot

While generative AI evokes images of omnipotent robots, its success hinges on the human beings that manage it. This holds true with conversational BI tools that apply generative AI to business intelligence and analytics. This blog defines conversational BI, why companies should consider it, and how their power users and casual users can best get the results they need. 

Generative AI (GenAI) creates digital content such as text, images, or audio after being trained on a corpus of existing content. The most popular form of GenAI centers on a language model (LM), which is a type of neural network that interprets, summarizes, and generates text. BI vendors such as ThoughtSpot are adding conversational BI capabilities to support tasks such as the following. 

  • Data preparation. Prepare data for analytics by filtering, enriching, cleaning, formatting, and structuring it. 

  • Cataloging. Organize metadata such as file names, database schemas, and table attributes to assist in data discovery, assessment, and governance.  

  • Analytics. Analyze data and visualize the outputs in reports and dashboards, or feed them into applications, workflows, and alerts. 

Practitioners like the concept. In fact, 40% of 236 respondents to a recent BARC/Eckerson Group survey are experimenting with, implementing, or using conversational BI. Another 38% are discussing or evaluating it. Adopters like these seek faster time to insight, reduced workloads, and simpler access to BI outputs. And they aim to reduce risks related to data privacy, output accuracy, and regulatory compliance. 

Conversational BI appeals to both power users and casual users of analytics.  

  • Power users—including data analysts, data scientists, data engineers, data stewards, and application developers—can gain productivity. Conversational BI helps them develop, debug, and document code, evaluate analytical outputs, and create predictive features.  

  • Casual users—including business managers that consume and explore data—can gain self-service. Conversational BI helps them generate or explain reports, models, and insights for them, and trigger application tasks based on those outputs. 

These groups have different expectations. The BARC/Eckerson Group survey showed that power users are somewhat pessimistic: 90% of them expect moderate to low or very low improvements in their use of BI and analytics. In contrast, business managers are optimistic: 77% of them believe GenAI will improve the use of BI in their company to a moderate, high, or very high degree. 


To what degree will GenAI improve the use of business intelligence and analytics in your organization in the next 12 to 18 months? 


This dichotomy underscores the need for distinct best practices with each group. Let’s examine what that looks like. Spoiler alert: you need humans in the loop throughout the process. 

Power Users and Casual Users

Power Users 

Power users can boost productivity and reduce governance risk by experimenting with conversational BI and teaching casual users how to use it responsibly. 

Experiment. As BI experts, power users experiment with new techniques to define what works and what doesn’t. They should prompt conversational BI tools to generate code for a wide range of tasks related to data preparation, querying, and analysis, then inspect the outputs. Which prompts work best for re-formatting or merging data? Does the tool generate accurate scripts for filtering or validating certain data sets but not others? As they answer myriad questions like these, power users can document which trustworthy prompts and artifacts can be reused. 

Teach. The resulting knowledge base enables power users to teach others. They teach their peers within BI teams how to reuse those trusted prompts and artifacts. More importantly, they should train casual users how to put them to work. An executive sponsor should reinforce this training by creating a center of excellence that showcases examples of successful projects by casual users. The more power users evangelize proven techniques, the better they empower casual users to serve themselves without creating contradictory analyses or other governance risks. They also can use the self-learning features of conversational BI tools such as ThoughtSpot that adapt their outputs based on user feedback—which then guides casual users appropriately. 

Casual Users 

Casual users can achieve self-service and reduce governance risk by scoping their business objectives and learning from power users. 

Scope. As business experts, casual users should scope the business problems that conversational BI might help solve. For example: 

  • To meet holiday season deadlines, a supply-chain manager might need to make intra-day adjustments to product orders or warehouse inventories based on changing factors such as shipment status, inclement weather, and exchange rates.  

  • To optimize pricing in dynamic markets, the owner of an insurance product line might need to toggle between many ad-hoc views of how various market variables correlate to one another. 

Casual users such as these often cannot get the real-time insights they need from their own traditional BI license. They must describe their business problems to power users in detail so that power users can scope how conversational BI can help. 

Learn. In addition, casual users must learn proven techniques from power users. For example: 

  • The supply-chain manager might learn that her natural-language prompts must contain certain product details and reference data to get accurate results. She also might learn to rely on a human power user rather than conversational BI when analyzing a certain product line. 

  • The insurance manager might learn that the conversational BI tool generates accurate correlations of market variables in North America and Europe, but not Asia. He also might learn to swap in a different language model—say Google’s Gemini rather than OpenAI’s ChatGPT—to compare certain customer groups. 

Casual users must become champions of what they learn, helping their peers align with best practices and scope additional business problems for power users.  

Summary 

Power users will reduce their backlogs by experimenting with conversational BI, then teaching casual users how to serve themselves. And casual users that understand and communicate their objectives can learn how to do this without incurring significant new risk. By creating this virtuous feedback loop, these teams can collaborate to harness the power of GenAI and put the omnipotent robots to good use. To learn more about conversational BI, be sure to check out our Eckerson Group report on this topic. 

Kevin Petrie

Kevin is the VP of Research at BARC US, where he writes and speaks about the intersection of AI, analytics, and data management. For nearly three decades Kevin has deciphered...

More About Kevin Petrie