With a keen eye for data science and a vision for the future of business intelligence, Chloe Maraina is an expert at translating complex data into compelling visual stories. Today, she joins us to unravel the profound impact of agentic artificial intelligence on the analytics landscape. We’ll explore how this technology is fundamentally reshaping BI from a tool for creating reports into a dynamic, intelligent workflow. Our discussion will cover the practical differences end-users will experience, how AI agents are demolishing long-standing bottlenecks in modeling and dashboarding, and what truly sets a coordinated, multi-agent system apart from a single AI assistant. Finally, we’ll look toward 2026 and beyond, envisioning a future where businesses can build their own custom AI agents to solve unique challenges.
The article highlights a shift to an “agentic system” to treat analytics as a “workflow rather than just a means of creating analytic artifacts.” Could you walk us through the practical, step-by-step differences an end-user will experience with this new workflow-based approach?
Absolutely. The difference is night and day; it’s like moving from a manual assembly line to a fully automated one. In the old world, if a marketing manager wanted to understand campaign performance, their journey was fragmented and slow. They would submit a request, wait for a data analyst to become available, who would then manually write queries, join data tables, and painstakingly design a dashboard. It was a series of handoffs, with each step creating delays. Now, with an agentic workflow, that same manager can simply ask, “How did our latest campaign impact sales in the western region?” Instantly, a team of specialized AI agents gets to work. One agent understands the natural language question, another validates the underlying data model, and a third, like SpotterViz, generates a series of relevant visualizations on the fly. The experience becomes a seamless, interactive conversation with your data, completely eliminating those frustrating waiting periods and transforming analytics from a static report into a dynamic, real-time discovery process.
You’ve identified modeling and dashboarding as key bottlenecks where “enterprises lose days rather than minutes.” Can you share an anecdote or a specific metric that illustrates how SpotterViz or SpotterModel will dramatically reduce that cycle time for a typical analytics team?
I vividly recall a project with a retail client where the analytics team was tasked with building a comprehensive model to connect inventory, sales, and supply chain data. It was an incredibly talented team, but they spent nearly a week just on the modeling piece—ensuring data governance was right, that the semantics were correct, and that everything was accurate. After that, they spent another few days just on the dashboard, meticulously arranging charts and making sure the story was clear. The entire process was a multi-week ordeal. With tools like SpotterModel, that initial week of modeling could be compressed into a few hours. An analyst can now instruct the agent to build the semantic model using natural language, and the AI handles the complex, code-free assembly. Similarly, SpotterViz takes the dashboarding process from days of manual layout work to mere minutes of AI-assisted generation. The shift is monumental; we’re literally collapsing a process where enterprises lose entire days of productivity into something that takes less time than a coffee break.
The content contrasts a coordinated multi-agent system with tools like Power BI’s Copilot. Beyond just having multiple agents, what specific collaborative tasks can these Spotter agents perform together that a single AI assistant cannot? Please provide a detailed example of their interaction.
This is a fantastic question because it gets to the heart of the architectural advantage. Think of a single AI assistant as a highly skilled generalist, a jack-of-all-trades. It’s powerful, but it tackles tasks sequentially. A multi-agent system, on the other hand, is a team of collaborating specialists. Let’s imagine a complex business question: “Forecast our Q4 revenue based on current sales trends, but factor in the potential impact of the upcoming marketing promotion.” A single assistant would have to process this step-by-step. A multi-agent system works in parallel. The primary agent first deconstructs the request. It then tasks SpotterModel to access and validate both the historical sales data and the promotional forecast data. Simultaneously, it engages SpotterViz to begin mocking up potential visualization formats—like a forecast chart with confidence intervals. As soon as the model agent confirms the data is integrated and sound, it signals the viz agent, which instantly populates the pre-designed chart. This coordinated handoff and parallel processing, where one agent is preparing the canvas while another prepares the paint, is something a single assistant simply can’t do. It’s this collaborative workflow that provides a deeper, faster, and more reliable answer.
Looking ahead to 2026, when users can build their own agents, what specific business problems or unique company workflows do you envision customers addressing with these custom agents, and what technical skills might be required for a non-developer to build one successfully?
The potential here is enormous because every business has its own unique operational rhythm and challenges. I envision a manufacturing company creating a “Predictive Maintenance Agent.” This custom agent would be trained to monitor real-time sensor data from factory equipment, cross-reference it with historical maintenance logs, and proactively schedule repairs before a breakdown occurs, saving millions in downtime. A financial services firm could build a “Compliance Monitoring Agent” that constantly scans transactions and communications for regulatory red flags. The beauty is that building these won’t require a computer science degree. The interface will be low-code or no-code, focusing on business logic. A line-of-business manager would define the agent’s objective, point it to the necessary data sources, and establish a set of rules and actions using a simple, guided interface. The required skills will shift from coding to a deep understanding of the business process you want to automate. It’s about being an expert in your business, not an expert in programming.
What is your forecast for agentic AI in business intelligence? Will we see a complete shift from automating specific tasks to automating entire strategic decision-making processes for businesses within the next five years?
My forecast is one of powerful evolution, but not a complete abdication of human strategy. Within the next five years, we will absolutely see the full automation of complex operational decision-making processes. The days of an analyst spending a week to produce a single report to inform a decision will be gone. Instead, an agent will continuously monitor a business function, like inventory management, and not only surface insights but also recommend actions with clear, data-backed rationales. For instance, a “Decision Agent” might propose, “We should reallocate 500 units of Product X from the Midwest warehouse to the West Coast to meet projected demand, with an estimated 15% increase in regional profit.” However, the final, truly strategic decisions—like entering a new market or launching a transformative product line—will still require human oversight, intuition, and ethical judgment. The role of the human expert will be elevated from a data interpreter to a strategic arbiter, using the options and scenarios presented by AI to make the final, critical call. The agents will perfect the science of the decision, leaving us to master the art.
