KNOWIDEA V3.5: Moving to Agentic Business Intelligence

Date: January 26, 2026 To: All Stakeholders / Users From: Engineering & Product Team Subject: The Release of KNOWIDEA V3.5

We live in an era where AI models regularly outperform humans in global coding competitions and solve open problems in advanced mathematics. Given their intellectual capability, demonstrated in some of the most challenging tasks known to humankind, it seems obvious that these same models should be able to offer business intelligence insights that outperform any human analyst.

So, why hasn't that happened yet?

Why are most "AI for BI" tools still just fast but shallow wrappers?

The answer is architecture. Business intelligence requires more than just raw processing power; it requires specialized reasoning paths. A brilliant mathematician doesn't automatically make a great business consultant. They need to know where to look and how to frame the problem.

Today, with the release of KNOWIDEA V3.5, we are bridging that gap. We are moving beyond simple LLM calls to a multi-agent system designed to mimic the actual workflow of a professional analyst team.

We didn't just build faster agents; we built smarter ones. To create V3.5, we spent months consulting domain experts such as senior data analysts, strategy consultants, and BI veterans, to answer one fundamental question:

"What is the precise cognitive process you use to extract a 'good insight' from a raw dataset?"

We mapped their answers, heuristics, intuition, and verification steps, and codified them into our system. We taught the AI not just to read the data, but to follow the specific investigative paths that human experts use to uncover value.

Here is an overview of the architectural shifts and new features born from this research.

We realized that one model cannot do it all efficiently. You don't ask a research scientist to answer the phone, and you don't ask customer support to write a comprehensive white paper. V3.5 introduces two specialized agents, orchestrated by a routing layer that automatically directs user queries to the right expert.

The Pluto Agent (Deep Research):

When a user needs depth, Pluto takes over. Pluto is designed for complex reasoning and deep-dive analysis. In our internal blind tests, human analysts and consultants consistently selected Pluto's outputs over outputs from out-of-the-box reasoning models (including Claude 4.5 Sonnet, Gemini 3, and Deepseek V3.2). Pluto creates the kind of nuanced, high-level business insight that previously required a human associate.

The Mercury Agent (Instant Response):

Not every question needs a thesis. When a user needs a quick stat or a standard aggregation, such as calculating an average, max, or min, Mercury delivers instant answers with zero latency friction.

We have overhauled the UI to function less like a passive query tool and more like a daily intelligence briefing.

Visual & Narrative Synthesis: We have moved beyond isolated charts. Each insight card now combines a data visualization with a clear, journalistic headline and a Pluto-generated executive summary. This ensures users understand the story behind the numbers (e.g., the trade-off between revenue and retention) at a glance.

Powered by the Pluto agent, our report generation has graduated from "summary" to "analysis."

Depth over Brevity: The new reports are significantly more detailed. Instead of surface-level bullets, Pluto generates comprehensive narratives that connect dots between disparate data points, offering the level of detail expected from a specialized consultant.

We have removed the barrier between the user and their metrics. You no longer need to know SQL, Power BI, or navigate complex chart builders to create a monitoring deck.

Conversational Construction: Powered by Mercury, dashboard creation is now a conversation. Simply type your request into the command bar (e.g., "Visualize Relationship Manager revenue analysis"), and the agent intelligently selects the best visualization format, whether it's a granular data table, a comparative bar chart, or a high-level KPI card.

Generate, Verify, Pin: We introduced a verification layer to ensure trust. Instead of dumping data onto the screen, the system generates a live preview of the insight first. You confirm the visualization is what you intended, and then pin it directly to your canvas. It is a seamless "what you see is what you get" workflow.

Real-Time Sync: Once pinned, these widgets are not static screenshots. They remain connected to your data source, updating in real-time as your numbers change.

The biggest bottleneck in BI is always messy data. V3.5 introduces a dedicated Data Cleaning Agent.

Automated Hygiene: This agent acts exactly like a seasoned Data Engineer. It identifies inconsistencies, fixes schema errors, and standardizes formats automatically. It cleans the mess so the user can trust the math.

KNOWIDEA V3.5 represents more than just a feature update; it is a fundamental shift in how businesses interact with data. We are moving away from the era of passive "self-service" BI and into the era of active, agentic intelligence.

By codifying the workflows of human experts into Pluto and Mercury, we have built a system that doesn't just wait for your input; it acts. It cleans the mess, it investigates the anomalies, and it writes the reports.

We are proud to provide what we believe is the most powerful agent in business intelligence today. The barrier between having data and understanding it has never been lower.

We invite you to log in and meet your new intelligence team.

Welcome to KNOWIDEA V3.5

Status: Live Version: 3.5.0