IBM Patents an AI Agent That Runs Your Meetings From Start to Finish
IBM has filed a patent for an AI agent that doesn't just take notes in your meetings, it runs them. The system plans the agenda, talks to participants in real language, pulls in outside information, and writes the summary when it's done.

- Publication number
- US 2026/0300928 A1(US20260300928A1)
- Applicant
- INTERNATIONAL BUSINESS MACHINES CORPORATION
- Filed · Published
- Mar 25, 2025 · Oct 1, 2026
- Inventors
- Rui Han, Yuan Yuan Ding, Qi Liang Zhou, Zhong Fang Yuan, Heng Cheng Ke
- Status when we published
- Approved; patent expected soon
- Document
- 20 claims
- US classification
- 705/7.19
- Examiner
- BROCKINGTON III, WILLIAM S (Art Unit 3623)
What IBM's AI meeting host actually does for you
A conference room fills with people who don't know what the meeting is supposed to accomplish. That's a familiar problem, and IBM's patent describes an AI agent designed to step into the host role entirely.
The agent starts by reading a request (something like 'schedule a budget review with these four people next Tuesday'), then builds a full meeting plan on its own. During the meeting, it speaks with participants using natural language, pulls in relevant documents or data from connected tools, and keeps records updated as things progress.
When it's over, the agent writes a summary based on everything that happened. You don't assign someone to take notes or chase action items afterward, the AI handles the whole arc from scheduling through recap.
… generating, by the AI agent using a generative machine learning model, a meeting plan based on the request; and executing, by the AI agent, the meeting plan by: interacting with meeting participants using natural language processing …
Translation: The artificial intelligence uses a machine learning model to build and carry out a meeting schedule through natural conversation.
How the agent plans, talks, and summarizes in one loop
The patent describes a system built around a single AI agent, a piece of software that can take instructions, make decisions, and use other tools to complete a multi-step task on its own.
When someone asks the agent to host a meeting, it first assembles the basics: topic, participant list, and start time. It then uses a generative machine learning model (the same class of technology behind chatbots like ChatGPT) to draft a structured meeting plan, essentially an agenda with goals and expected talking points.
During the meeting itself, the agent does three things in parallel:
- Talks with participants using natural language processing (the ability to understand and respond to conversational speech or text)
- Queries external tools, calendars, databases, document repositories, or project management apps, to pull in relevant information on the fly
- Writes updates back to those same tools, keeping records current as decisions are made
After the meeting ends, the generative model produces a summary based on the full record of what was executed. The patent covers the method as a whole, not a specific interface or product form.
The AI agent executes the meeting plan by interacting with meeting participants using natural language processing, accessing relevant information from the one or more external tools, and updating meeting-related data using the one or more external tools.
Translation: During the call, the bot talks to participants, pulls data from external software, and updates records automatically.
What this means for workplace meeting software
Meeting software is already crowded with AI bolt-ons, transcription tools, auto-summaries, smart scheduling assistants. IBM's filing goes further by describing a system where the AI is the host, not a passenger. That's a meaningful difference: instead of assisting a human chair, the agent owns the meeting structure from the first invite to the last action item.
For enterprise software buyers, that framing matters. a growing pile of IBM enterprise-AI filings signals where the company sees agentic AI landing in corporate workflows. Whether this reaches a product is another question, but the direction, AI as a full participant rather than a passive recorder, is where a lot of enterprise tooling is heading.
IBM's 12th filing we've tracked since July in AI agents that act for you extends a pattern seen in one that fixes IT outages and one that manages conversation turns.
Source: US 2026/0300928 A1 at the USPTO (PDF). Full patent text and figures are in the official publication.
Get our take in your Top Stories
Liked this breakdown? Add Patentlyze as a preferred source on Google, and our plain-English take shows up more often in your Top Stories the next time IBM patent news breaks.
We read every patent application Big Tech publishes and send you the ones worth knowing. Plain English, free, every week.
The drawings · 8 sheets
8 drawing sheets from US 2026/0300928 A1 · click any drawing to enlarge

No comments yet