How to ask AI questions during a live meeting

There is a moment in most meetings where someone asks a question nobody can answer on the spot. What was the budget we agreed on for that campaign? Did legal already approve the revised terms? Who owns the integration with the new vendor? The conversation stalls. Someone promises to check afterwards. A follow-up email gets sent, maybe, three days later. By then, the context has cooled and the decision has drifted.

It does not have to work this way. If your notes, documents, transcripts, and decisions are already organised in one place, an AI assistant can search across all of them while you are still in the conversation. Not after. Not later. Right there, in the meeting, while the question is still warm.

Fabric's AI assistant is built for exactly this kind of moment. It sits on top of your entire knowledge base and responds to natural-language questions with answers drawn from your own content. During a live meeting, you can ask it what was decided last quarter, pull up a specific document, check a figure, or surface a thread from Slack that is relevant to what is being discussed. The assistant draws on Fabric's memory engine, which means it does not just match keywords. It understands context, recalls relationships between pieces of information, and retrieves what is relevant even when you phrase your question differently from how the original content was written.

Setting the stage before the meeting starts

The most useful thing you can do before a meeting is make sure the material you might need is already inside Fabric. This could mean uploading a few documents, connecting your Google Drive or Slack through Fabric's connected apps, or simply confirming that your previous meeting notes are saved. Fabric's cloud storage holds everything in one place, so there is no need to worry about which folder something ended up in or whether a file is on your laptop or in a shared drive somewhere.

If you have been using Fabric for your meeting notes, much of this groundwork is already done. Notes from previous meetings, transcripts, action items, and decisions all live in the same knowledge base. The AI assistant can reach into any of it. So when someone asks "What did we decide about the vendor shortlist last month?", the answer is already there, waiting to be retrieved.

You do not need to prepare a briefing document or re-read old notes yourself. The assistant does the retrieval. Your job is simply to ask.

How to query the AI assistant during a live call

Once you are in a meeting, querying Fabric is as simple as typing a question. You can ask things like "What is the current budget allocation for Q3?" or "Show me the last update from the design team on the rebrand." The assistant uses semantic search to find what you need, which means it looks for meaning rather than exact word matches. If the budget was discussed in a transcript and the number appeared in a shared spreadsheet, the assistant can surface both.

This works alongside meeting recording. You can have Fabric capturing the current conversation while simultaneously querying it about past conversations. There is no conflict between the two. You are not choosing between recording and searching. You are doing both, which means the current meeting becomes part of your searchable knowledge base as soon as it ends, ready to inform the next one.

A few types of questions that tend to be particularly useful mid-meeting: checking previous decisions ("Did we agree to delay the launch or keep the original date?"), verifying figures ("What was the conversion rate we reported in the September review?"), finding ownership ("Who was assigned the API documentation task?"), and pulling up supporting material ("Is there a brief for the new onboarding flow?"). These are the kinds of questions that typically derail meetings or generate follow-up work. Answering them on the spot keeps the discussion moving.

Using ChatGPT or Claude to query Fabric during a meeting

One of the more useful patterns is querying your Fabric knowledge base from within ChatGPT or Claude, using MCP (Model Context Protocol). This means you do not have to switch tools or leave the AI environment you are already comfortable with. If you are someone who keeps a ChatGPT or Claude window open during meetings for drafting responses or thinking through problems, you can now point that same interface at your Fabric content.

Through MCP, your preferred AI model gains access to your stored documents, notes, and meeting history. You can ask Claude to look up what was discussed in last week's product sync, or ask ChatGPT to find the slide deck that was shared before the board meeting. The AI retrieves the content from Fabric and presents it in the conversation, so you never have to break your workflow.

This is particularly helpful for people who use AI assistants as a kind of thinking partner during meetings. Rather than limiting the AI to its general training data, you give it access to your specific organisational context. It knows what your team has discussed, what decisions have been made, and what documents exist, because all of that lives in Fabric.

Giving AI scoped access with agent permissions

When you connect an external AI to your Fabric content, you probably do not want to hand over the keys to everything. Fabric's agent access lets you define exactly what a given AI session can see. You can scope access to specific folders, tags, or content types, and you can set time limits so that access expires after the meeting or after a defined window.

This matters for two reasons. First, it is a sensible security practice. Not every AI query needs access to your entire archive, and limiting scope reduces the surface area for anything unexpected. Second, it makes the AI more effective. When the assistant is working with a focused set of content, its answers tend to be more precise because there is less noise to filter through.

For a recurring meeting, you might set up a scoped agent that has access to all notes and documents tagged with that project. For a one-off review, you might grant temporary access to a specific folder. The flexibility is there to match the way you work, rather than forcing you into a single model of access control.

Combining live queries with meeting notes

If you are already using Fabric to take meeting notes without a bot joining the call, the AI assistant adds another layer. You can record and transcribe the current meeting with Fabric's voice notes, and at the same time query the assistant about content from earlier meetings.

This creates a useful loop. The notes from today's meeting flow into your knowledge base, where they become searchable material for tomorrow's meeting. Over time, your organisational memory becomes richer and more interconnected. Questions that once required someone to dig through emails or scroll through old documents can be answered in seconds.

You can also combine your meeting notes with the transcript after the meeting, creating a single record that captures both the raw discussion and the structured takeaways. When you query the assistant in a future meeting, it can draw on both, giving you the specific quote from the transcript or the summarised decision from the notes, depending on what you ask for.

Turning answers into actions

Finding information is only part of what makes this useful. The other part is what you do with it. If the assistant surfaces a decision from a previous meeting, you can confirm it on the spot and move the conversation forward. If it finds a gap (perhaps no decision was recorded, or the action item was never assigned), that becomes visible immediately rather than weeks later.

Fabric can also turn meeting notes into action items automatically, which means the outputs from your current meeting feed directly into task lists and follow-ups. The combination of live querying and automatic action-item generation means that meetings become both better informed and more productive. You spend less time searching for context and less time writing up what happened afterwards.

Security and trust

Querying your knowledge base during a meeting means sensitive information may be surfaced and displayed on your screen. Fabric handles this with AES-256 encryption and CASA Tier 2 compliance, which means your content is protected both at rest and in transit. The scoped agent access described above adds another layer, ensuring that AI queries only reach the content you have explicitly made available.

This is worth considering if you work in an environment where confidentiality matters, which is most environments. The ability to ask an AI about previous discussions and decisions is only useful if you trust the system holding that information. Fabric is built as a secure gateway to your content, not a platform that exposes it.

Frequently asked questions

Can I query the AI assistant while a meeting is being recorded?

Yes. Recording and querying are independent functions. You can have Fabric capturing audio and generating a transcript while simultaneously asking the AI assistant about content from earlier meetings or documents. The two do not interfere with each other.

Does the AI assistant only search meeting notes, or can it access other content too?

The assistant searches your entire Fabric knowledge base, which includes meeting notes, transcripts, uploaded documents, content from connected apps like Google Drive and Slack, and anything else you have stored. The memory engine draws connections across all of this content, so you are not limited to a single type of source.

How does semantic search differ from regular keyword search?

Semantic search understands the meaning behind your question rather than just matching words. If you ask "What was the budget for the new hire programme?" it can find relevant content even if the original document used terms like "recruitment allocation" or "headcount spend." This makes it much more forgiving of how you phrase things, which is helpful when you are asking questions quickly during a meeting.

Can I use ChatGPT or Claude to access my Fabric content during a meeting?

Yes. Through MCP, you can connect ChatGPT or Claude to your Fabric knowledge base and query it from within those tools. This means you can use whichever AI assistant you prefer without switching between applications. The AI retrieves content from Fabric and presents it in your existing conversation.

What is agent access, and how does it work during meetings?

Agent access lets you give an AI model scoped, time-limited access to specific parts of your Fabric content. You define what the agent can see (certain folders, tags, or content types) and how long the access lasts. This is useful during meetings because you can grant access that is relevant to the discussion without opening up your entire knowledge base.

Is my data safe when I query the AI assistant?

Fabric uses AES-256 encryption and holds CASA Tier 2 certification. Your content is encrypted at rest and in transit, and agent access is scoped so that AI queries only reach the content you have made available. The system is designed as a secure gateway, not an open pipe to your data.

Do I need to prepare anything before the meeting for this to work?

The main requirement is that your relevant content is already in Fabric. If you have been using Fabric for meeting notes and document storage, or if you have connected apps like Google Drive and Slack, the content is already there. No special preparation is needed for individual meetings beyond ensuring your knowledge base is reasonably up to date.

Can I ask the AI assistant about decisions made in previous meetings?

Yes, this is one of the most common use cases. If your previous meeting notes and transcripts are stored in Fabric, the assistant can search across them to find specific decisions, action items, figures, or discussion points. You can ask questions like "What did we decide about the pricing model in the August review?" and get an answer drawn from the relevant notes.

Does the AI assistant work with content from connected apps?

Yes. If you have connected Google Drive, Slack, or other apps through Fabric's integrations, the AI assistant can search across that content as well. This means you can ask about a document stored in Google Drive or a conversation thread from Slack without leaving your meeting or opening those apps separately.

Can multiple people on the call query the AI assistant at the same time?

Each person queries their own Fabric knowledge base independently. If multiple team members are using Fabric, each can ask their own questions during the meeting without affecting the others. The answers each person sees are drawn from their own stored content and whatever has been shared with them.

Related pages

• How to take meeting notes without a bot joining the call

• How to combine your meeting notes with the transcript

• How to turn meeting notes into action items automatically