1. Introduction — The Evolution of Analytics Over the couple of decades, since the dot com & e-commerce boom followed by the advent of bigdata, analytics has mainly revolved around SQL queries, excel pivot tables and data visualizations in for of charts and dashboards.  Analysts have been the primary bridges between data and decisions, crafting queries, building visualizations, and providing visibility for the business leadership to help informed decision making. But as language models evolve and meet structured data, a new paradigm is emerging: agentic AI where the AI systems are capable of querying, summarizing, validating and publishing insights autonomously. Instead of clicking around and refreshing dashboard filters, teams can now converse with their data in natural language, for example asking “Why did our retention drop month over month?” or “How does the Q4 revenue forecast look like, and what will be the major drivers?” 2. The Old Paradigm — Manual Querying and Passive Dashboards Traditional BI tools are were designed to be reactive. Analysts building visualizations in for of dashboards powered by complex SQL queries where each new iteration can take sprints of work slowing down business’s decision-making process as they rely on up-to-date insights. These tools are great in determining what happened along with the detection trends but often fall short in the next level of deep dive analysis to answer why something took place and as a result the bottleneck falls on the analysts to surface those answers. 3. The Shift — Agentic AI on Top of Trusted Data Layers The future of analytics is self-serve and conversational. By layering agentic AI on top of a semantic data layer (a governed repository of metrics and logic), organizations enable AI agents to help business users find greater insights in their data by asking deeper questions in natural language and helping agents get more accurate by improving their knowledge base through feedback loop. This process allows business users to find quick answers to all routine and what if scenarios, therefor freeing up the analyst’s time solving complex critical problems or build more agents to serve various workflows. The performance of the agent would be defined by the sematic layer powering the agentic AI application, as well as the precise prompts and knowledge base to setup the agents for example information about all data tables related to business problem, business context definitions, metric definitions, related dimensions as well as final output formatting, style and tone. Example prompts: “Business wants to look at all the top performing products (in terms of net sales) in the past year broken down by quarter, in each geographical region. Also perform an analysis on product type and category in each region. Create summary stats for top 5 highest and lowest performing products in each region, and compare it year over year” “Analyze trends in total revenue and customer signups over the past 6 months, and analyze cancellation rate or churn, also summarize the top 5 reasons for cancellations in North America and Europe?” 4. The Benefits — From Passive to Proactive Analytics Working with an agentic AI helps turbocharge the curiosity nerve as the Agentic AI doesn’t just answer, it actually thinks ahead and helps you dive deep into data patterns you may have not anticipated before within seconds. Pairing this on top of your existing dashboards makes the analysis even more powerful. The agent also relies on your feedback of correct responses to further improve its accuracy. You may consider the agent to be a junior analyst who always curious to learn, while always available to crunch the most complex analysis. 5. The Enablers — Building the Foundation for Agentic Analytics To make this vision realistic, data teams must focus on these three aspects: Foundational Semantic Metric Layer: This is the single source of truth where all business logic lives. This may contain metric aggregation on the top of existing normalized facts and dimensions.
Robust data quality checks : The high level of automated data quality checks on the semantic later data assets would be instrumental for AI to provide highly accurate trustworthy responses and avoid any hallucinations
Conversational Interface Layer : The natural language responses that can be verified to the source, e.g. cited design document, confluence site, team’s charter, data sources, vetted queries etc. This would not increase trust in the data but also allow the user to trace AI responses directly form the source for further deep dive. Foundational Semantic Metric Layer: This is the single source of truth where all business logic lives. This may contain metric aggregation on the top of existing normalized facts and dimensions. Robust data quality checks : The high level of automated data quality checks on the semantic later data assets would be instrumental for AI to provide highly accurate trustworthy responses and avoid any hallucinations Conversational Interface Layer : The natural language responses that can be verified to the source, e.g. cited design document, confluence site, team’s charter, data sources, vetted queries etc. This would not increase trust in the data but also allow the user to trace AI responses directly form the source for further deep dive. 6. Responsible Agentic AI: Guardrails for AI-Driven Insights Autonomous agents need governance and guardrails. Having robust data quality checks, source citations and human feedback loop to vet responses is essential for any AI system to provide accurate responses. There would still be a large human component involved for crosschecking sensitive financial statement data or medical summary data. The role of an analyst will not be obsolete but rather evolve to tackle more of strategic interpretation rather than redundant data wrangling that would be taken care by the agentic Ai system. Therefore, the analyst role would evolve from a traditional query writer, financial model builder and dashboard creator to a prompt engineer, data insights validator, data governance & quality steward. 7. Conclusion — The Future of Data Work We’re entering an era od data analysis beyond regular spreadsheets pivot tables and dashboards powered by agentic AI systems that are conversational in nature. Agentic Ai systems won’t eliminate analytics team, but rather enable them to do more and solve complex problems for the business. It would enable business to make decisions faster and deep dive into their data to discover new insights. The organizations that thrive in this age would be those who build foundational sematic data layers, have robust data quality & governance, have solid prompt engineering and feedback loop mechanism while continuously improving accuracy and trust in agentic AI systems. The next generation of analytics isn’t about faster loading dashboards; it’s more about intelligent and deeper conversations about your data.

This story contains new, firsthand information uncovered by the writer.

The is an opinion piece based on the author’s POV and does not necessarily reflect the views of HackerNoon.

Conversational Analytics: the Next Generation of Data Analysis and Business Intelligence

About Author

Comments

TOPICS

THIS ARTICLE WAS FEATURED IN

Related Stories

The Noonification: How Often Do NFTs Pass The Howey Test? (1/13/2023)

Darwin's Hybrid Intelligence to Align AI & Human Goals for Startups & VCs

The Noonification: White Man (11/26/2022)

The Noonification: The Metaverse is a Sh*tshow (11/2/2022)

100 Days of AI Day 1: From Newsletter to Podcast, Leveraging AI for Audio Transformation

10 Ways AI Has Changed Our Lives

The Noonification: How Often Do NFTs Pass The Howey Test? (1/13/2023)

Darwin's Hybrid Intelligence to Align AI & Human Goals for Startups & VCs

The Noonification: White Man (11/26/2022)

The Noonification: The Metaverse is a Sh*tshow (11/2/2022)

100 Days of AI Day 1: From Newsletter to Podcast, Leveraging AI for Audio Transformation

10 Ways AI Has Changed Our Lives

Light-Mode

Classic

Newspaper

Minty

Dark-Mode

Neon Noir

Minty

HN StartUps