Ongoing Research · Presentation Assistant

Your AI Co-Pilot for
Complex Meetings.

Business meetings across every sector rely on static slide decks built from stale data exports. We are building an AI agent that joins meetings as an invited third party — with live access to logs, dashboards, and data sources — so retrieval and explanation become seamless. 110ms for retrieval. 4 seconds for full LLM-powered explanation.

Scroll
Back to Research
Project Overview

The Problem We Saw in Business Meetings Across Every Sector

Every industry has complex dashboards — SCADA in energy, Bloomberg terminals in finance, CRM in real estate, ERP in manufacturing. But when teams meet to debrief operations, they export yesterday's data into static slides. By the time the presentation hits the screen, the numbers are already stale. The meeting becomes a ritual of reading outdated numbers aloud, not a forum for real-time decision-making.

The dashboards exist. The data exists. But the gap between what the dashboard knows and what the meeting participants can access in real-time is enormous. Someone asks a question, the presenter says “I'll check and get back to you.” The moment is lost. The decision is deferred.

We are building an AI agent that joins these meetings as an invited third party — with direct access to logs, dashboards, and data sources — so that any question can be answered instantly. Retrieval in 110 milliseconds. Full LLM-powered explanation in 4 seconds.

What We Are Researching

An AI agent that joins business meetings with live data access, replacing static slide decks with real-time retrieval and explanation.

110ms
Retrieval Latency
Data to screen
4 sec
LLM Explanation
Full answer generation
1 wk/mo
Time Wasted
On manual slide preparation
Any
Sector
Domain-agnostic deployment

Agent as an invited third party

The agent joins the meeting like a knowledgeable colleague. It has pre-configured access to the relevant dashboards, logs, and data sources. When someone asks “what happened with the western pipeline last Tuesday?” or “how did the portfolio perform against benchmark this quarter?” the agent retrieves the answer instantly and explains it in context.

Instant retrieval from any data source

110 milliseconds from question to data on screen. The agent connects to SCADA systems, financial databases, CRM platforms, ERP exports, log files, and any API-accessible data source. No more “I'll get back to you on that.” The answer is available in the moment the question is asked.

LLM-powered explanation in 4 seconds

Retrieval is only half the problem. The agent uses a large language model to explain what the data means in plain language, tailored to the meeting context. 4 seconds from data retrieval to a full, coherent explanation that everyone in the room can understand and act on.

Domain-agnostic, sector-independent

Energy, finance, real estate, manufacturing, logistics, healthcare — any sector that runs complex operations with complex dashboards. YAML configuration files define data sources, intent patterns, and dashboard mappings. No code changes needed for new domains.

The Market Gap

Every sector has complex dashboards. None has an AI agent that can join a meeting and answer questions about them in real-time.

Dashboards are everywhere. Meeting intelligence is nowhere.

Organizations invest billions in SCADA, ERP, CRM, BI tools, and data platforms. But the meeting room — where data meets decisions — remains stuck in static slides exported from those systems. No tool bridges the gap between what the dashboard knows and what the meeting can access in real-time.

Existing tools solve the wrong problem

PowerPoint, Keynote, Google Slides — all assume static content. AVEVA, GE, Tableau, Power BI — all assume a single user at a workstation. None assume a meeting where multiple stakeholders need to ask questions and get instant answers from live data across multiple systems.

The technology is mature

Browser-based retrieval at 110 milliseconds is viable. LLM-powered explanation at 4 seconds is standard. WebSocket real-time streaming is infrastructure. API access to enterprise data sources is ubiquitous. The pieces exist. No one has assembled them into a meeting intelligence agent.

The pain is universal across sectors

Energy operators debriefing production. Financial teams reviewing portfolio performance. Real estate firms assessing property portfolios. Manufacturers analyzing production lines. Logistics companies tracking shipments. Every sector has the same problem: complex dashboards, static meeting presentations, and no way to answer questions in real-time.

Design Principles

Four principles that shape the research.

Invited, not imposed. The agent joins the meeting as a trusted third party. It is configured with access to the specific data sources the meeting requires. It does not impose its own interface — it works within the existing meeting format, whether that is a screen share, a video call, or a conference room.
Retrieval first, explanation second. 110 milliseconds to retrieve the data. 4 seconds for the LLM to explain it. The two-stage architecture ensures that the meeting always has the raw data instantly, even if the explanation needs a moment longer.
Domain-agnostic, sector-independent. The same agent that debriefs an energy operations meeting can debrief a financial portfolio review or a manufacturing production report. YAML configuration defines the data sources, intent patterns, and dashboard mappings for each domain.
Same pipeline, different domain. The same five-stage intelligence pipeline that powers our procurement, insurance, and mobility research, adapted to the meeting room. Same engine, different data, different sector.
Plan Moving Forward

Where This Goes Next

The research is complete. The next phase is pilot deployment with organizations that run complex operations meetings.

01
Pilot with a complex-operations team. Deploy the agent in live meetings — energy operations briefings, financial portfolio reviews, manufacturing production reports. Measure time saved, decision speed, and question resolution rate.
02
Expand across meeting types. Move from operational debriefs to strategy reviews, compliance audits, investor presentations, and cross-functional briefings. Each use case validates the agent's domain-agnostic architecture.
03
Open-source community. Release the platform as open-source. Build a community of practitioners who contribute data source connectors, intent patterns, and domain configurations for their sectors.
04
Cross-sector deployment. The meeting intelligence problem is universal. Energy, finance, real estate, manufacturing, logistics, healthcare — every sector that runs complex operations with complex dashboards needs this. The platform extends naturally to all of them.

This Research Needs the Right Partners

The technology is ready. The pain is universal across sectors. We need organizations that run complex operations meetings and want to pilot an AI agent that joins as an invited third party.

Explore the Laboratory

See the systems

Explore the operational and financial environments we are developing.

Explore Systems →

Examine the projects

See how individual problems become working intelligence systems.

View Projects →

Explore autonomous agents

See how domain-specific agents operate inside defined workflows.

Explore Agents →

Discuss a partnership

If your organization supports research, technology development, or infrastructure, there may be an opportunity to contribute.

Start a Conversation →