We’re moving from a search-and-click economy to a question-and-answer economy.

Since the inception of Business Intelligence (BI) tools and dashboards, data analysts and users of these tools, for the most part have followed a familiar usage pattern: identify dashboard > open dashboard > apply filters to narrow dataset > and then finally (in a lot of cases) get to the single insight of interest or action. For example, a business analyst might be interested in a revenue variance for a specific period from their business unit, rather than the entire time series of monthly revenue across all business units. Importantly, the goal is not simply to explore the data, but to drive a decision or a call-to-action.

There is a case to be made that the route-to-insight could be shortened, made quicker and more intuitive. To be clear, traditional dashboards will remain and still serve as a canvas to provide overviews of KPIs/trends visuals and information. However, AI-enabled systems will increasingly play an important role in driving how BI data is used for gathering quick insights and trends, find anomalies in their data and self-serve as an automated route to faster decision making. In short, we are entering into the era of the AI/BI Dashboards, where this tools will be used as Q&A assistants not just search-and-click interfaces.

AI/BI dashboards will be part of what I see as conversational flow systems alongside web apps, data agents or chatbots ¹. There will be an embedded automation layer that also uses the outputs of Large Language Models LLMs to deploy automated reporting and trigger agent workflows. 

¹ AI system architecture

We see this already with Databricks AI/BI Genie app ², a web app which when prompted by users in natural language, uses a compound AI system (mix of LLMs) to provide a text and visual summary of the datasets in its data model/catalogue. Databricks describes it as a conversational user experience which allows the users to interact fully with the data, create and share prompt outputs.  The Genie app is also embedded in Databricks native BI dashboards to enable you to write prompts to generate data.

² Databricks Genie app (source: databricks.com)

This is similar to Microsoft’s Copilot in Power BI or their Fabric data agents which offer conversational Q&A systems using generative AI ³ [Ref]. Fabric data agents also enable agent-to-agent automation that lets multiple agents share goals, memory, and reasoning context. This allows multi-agent AI solutions which are increasingly popular, thanks to no/low code platforms like n8n and Zapier, work together more effectively and deliver richer, more complete responses and complex automations.

³ Microsoft Fabric Agent (source: microsoft.com)

Conversational BI – where the true value lies

Regardless of the AI/BI tool used, I believe their true value will come from workflow automation and reporting, and hypothesis driven multi-step Q&A workflow. Taking the first of these, a financial business analyst may be interested in quickly searching for exceptions or anomalies in their reports without directly accessing a dashboard, or perhaps where late payments may come from or who is the largest debtor? These AI-enabled systems will generate daily summarised emails and reports and perform predictive analysis reports (which bills are likely to be late). 

Multi-step Q&A prompt capabilities in e.g. Databricks’ Deep Research mode ⁴, allows users to sketch out their reasoning process to a business problem and save the instructions as ready-made prompts, effectively going from “how” and “what” type questions to “why”. Think of it as going from “I know you can tell me what my revenue for the month is, but tell me why they are down on budget, and by the way these are reports I reference and these are reasons I check them for context“.

This kind of layered, contextual questioning transforms BI into a strategic partner and not just a reporting tool.

⁴ Databricks BI Genie Deep Research Reasoning (source: databricks.com)