Meeting Topics and Categorization

Every meeting in Agency Hero produces more than a transcript. As soon as processing finishes, the AI breaks the conversation into discrete topics — the specific themes you discussed — and maps each one to your workspace’s shared topic registry. Over time, those topics become the organizing spine of everything your workspace knows: decisions, action items, risks, questions, and AI-generated briefs all connect through them.

This article explains how topics work at every layer — from how they’re detected in a single meeting to how they accumulate into workspace intelligence over time.

What Meeting Topics Are#Copy link

A meeting topic is a named theme extracted from a meeting’s content. It represents a coherent thread of conversation — something like “API Migration”, “Q3 Budget”, “Onboarding Timeline”, or “Competitive Positioning” — that the AI identified as a meaningful subject within that specific call.

Topics exist at two levels in Agency Hero:

• Meeting-scoped topics — The individual themes extracted from a single meeting. These are the raw output of post-meeting AI analysis. Each meeting has its own set of them, ranked by how much discussion time and significance they received.
• Workspace-level topics (the canonical registry) — The shared list of recurring themes for your workspace. When a meeting-scoped topic is processed, Agency Hero tries to match it to an existing canonical topic. If it matches, that meeting’s contribution is added to the ongoing record. If nothing matches, a new canonical topic may be created.

This two-level structure is what makes topics powerful: a single meeting topic is just a data point, but the canonical workspace topic it maps to is a living longitudinal thread — one that accumulates every time that theme appears across your meetings.

How Topics Are Automatically Detected#Copy link

Topic detection happens automatically as part of the post-meeting processing pipeline, which runs after a meeting transcript is confirmed. The process has three steps:

1. Topic extraction#Copy link

The first AI agent reads the full transcript and identifies the discrete themes discussed. This produces a list of meeting-scoped topics — each with a name, a short summary of what was said, a list of key points, and a rank reflecting how central it was to the conversation.

2. Canonical matching#Copy link

Once meeting-scoped topics are extracted, the system runs a multi-step topic matching pipeline against your workspace’s canonical topic registry:

• Exact / fuzzy match — The topic name is normalized (lowercased, whitespace-stripped) and compared against existing canonical topics. Close text matches are resolved first.
• Semantic / embedding match — If no close text match is found, the system computes a vector embedding of the topic and compares it against the embeddings of existing canonical topics using cosine similarity. This catches cases where the same theme appears under slightly different phrasing (e.g., “Infrastructure Costs” and “Cloud Spend”).
• High-confidence auto-link — If the match exceeds the confidence threshold, the meeting topic is automatically linked to the matching canonical topic.
• Proposed link (review required) — If the match is plausible but not certain, Agency Hero proposes the match and surfaces it for your review during the post-meeting workflow. You can confirm or redirect.
• Create new — If no existing canonical topic is a good match, the system can propose creating a new one. You can accept the proposal (which registers the new topic in your workspace) or redirect the meeting topic to an existing one instead.

3. Linking intelligence items to topics#Copy link

Every action item, decision, risk, and open question extracted from the meeting is linked to its primary topic. This is what allows the canonical workspace topic to accumulate confirmed intelligence over time — confirming an action item from a meeting automatically feeds the relevant canonical topic.

Viewing Topics on an Individual Meeting#Copy link

After a meeting is processed, you can see its topics in the Meeting Items tab of the post-meeting review panel.

Topics section (ranked by importance)#Copy link

Topics are listed in ranked order based on how much discussion time and significance each received during the meeting. Each topic card shows:

• Rank badge and topic title
• Topic details — A short summary of what was discussed under this theme
• Key points — The specific claims, decisions, or items that came up within the topic
• Intelligence items — The action items, decisions, risks, and open questions nested under this topic
• Workspace topic link — A badge or tag showing which canonical workspace topic this meeting topic has been matched to (if matched)

The workspace topic link is especially useful — clicking it takes you directly to the canonical topic page in Knowledge → Topics, where you can see the topic’s full history across all meetings.

General Items (not topic-specific)#Copy link

Some extracted items don’t naturally belong to a single topic — for example, a standalone commitment made at the end of a meeting that doesn’t fit cleanly into any discussed theme. These appear in the General Items section of Meeting Items, grouped by type (action items, decisions, risks, open questions) rather than by topic.

Confirming, Editing, and Managing Topics on a Meeting#Copy link

During the post-meeting review, you have full control over how topics are handled:

Review proposed topic matches. When the AI proposes linking a meeting topic to a canonical workspace topic, you’ll see a suggestion badge with a confidence indicator. You can:

• Confirm the link to accept the match
• Reject the link if it’s wrong
• Change the link to reassign the meeting topic to a different canonical topic
• Unlink the meeting topic from any canonical topic entirely

Redirect a meeting topic. If you don’t agree with a proposed new topic, or want to map the theme to a different existing canonical topic, use the overflow menu on the topic card to change the link or merge into another topic.

Inline editing. Topic names and summaries can be edited in place from the meeting detail view. Changes save on blur or Enter.

Transcript references. Each topic card includes timestamp badges that link back to the exact moments in the transcript where that theme was discussed — useful for verifying the AI’s interpretation.

The Difference Between Meeting-Scoped Topics and Workspace Topics#Copy link

It’s worth being explicit about this distinction, since both are called “topics”:

When a meeting-scoped topic is linked to a canonical workspace topic, it becomes a contribution to that workspace topic — adding to its history, feeding its brief, and expanding its coverage.

How Topics Accumulate Over Time#Copy link

The real value of topics emerges over multiple meetings. Each time a canonical topic is touched by a processed meeting, three things happen:

1. The meeting is added to the topic’s provenance timeline. The workspace topic records which meetings have contributed to it, when, and what was discussed.
2. Confirmed intelligence items feed the topic. Any action item, decision, risk, or open question that is confirmed during review and linked to the topic is stored as part of the workspace’s intelligence ledger under that topic.
3. The Topic Brief is updated. The canonical topic has a living AI-generated brief that synthesizes everything linked to it. This brief is regenerated from three sources:
   • Meeting contributions — The most recent meeting topics linked to this canonical topic (up to 10), including their text, key points, and timestamps
   • Intelligence items — All confirmed items (decisions, risks, action items, open questions) tagged to this topic
   • Linked memos — Any workspace memos that have been associated with this topic

Topic Briefs#Copy link

The Topic Brief is the primary way you understand a topic’s current state at a glance. Navigate to Knowledge → Topics and select any topic to see its brief. The brief answers questions like:

• What’s the current status of this theme across the workspace?
• What decisions have been made?
• What’s still open or unresolved?
• What commitments exist?

Topic briefs are always derived, not manually authored — they’re generated from the underlying confirmed intelligence, so they stay current automatically as new meetings are processed and items are confirmed. You can also manually trigger a regeneration from the topic detail page.

Topics are also the connecting tissue for workspace AI features. When Agency Hero assembles a context pack to power AI chat, workflow agents, or meeting prep, it pulls the canonical topic list and their briefs as core context — making topics one of the highest-leverage inputs to AI quality across the workspace.

Pre-Creating Topics Before a Meeting#Copy link

One of the most impactful things you can do for a new workspace is create topics before the first meeting is processed.

Topics are the organizing structure the AI uses to route intelligence. When a meeting is processed and the system finds existing canonical topics that match what was discussed, it can make high-confidence automatic links — resulting in cleaner topic assignments and fewer items needing manual review. When the workspace has no topics yet, every theme extracted from a meeting becomes a proposal for a new topic, and the workspace starts from scratch rather than building on established structure.

How to create a topic manually#Copy link

1. Navigate to your workspace’s Knowledge → Topics tab.
2. Click New Topic to open the creation dialog.
3. Enter the topic name and confirm.

The topic is immediately registered in the canonical registry with a discussion count of zero. It will be available for AI matching on the next processed meeting.

What topics to create upfront#Copy link

You don’t need to be exhaustive or perfectly precise — the AI will still extract themes from the transcript regardless. The goal is to give the system recognizable anchor points. Good starting topics reflect:

• The main workstreams or deliverables you’re working on with this client (e.g., “Website Redesign”, “Q3 Campaign”, “Technical Integration”)
• Recurring business concerns likely to come up repeatedly (e.g., “Budget”, “Timeline”, “Stakeholder Alignment”)
• Known risks or open threads you’re already tracking (e.g., “Third-Party Dependencies”, “Legal Review”)

Even a rough set of 4–8 topic names gives the AI strong anchors to work with. You can always add, rename, merge, or archive topics as the workspace evolves.

Managing Your Workspace Topic List#Copy link

Your canonical topic list lives at Knowledge → Topics in the workspace. Common management actions:

• Create — Click New Topic, enter the name, and confirm. The topic is registered immediately.
• Rename — From the topic detail page, update the name directly. The topic is re-normalized for matching. If the new name collides with an existing topic, Agency Hero will prompt you to merge instead.
• Merge — Over time you may end up with two topics covering the same thread under slightly different names (e.g., “API Integration” and “Integration Architecture”). Use Merge into another… from the topic’s overflow menu to combine them. The winner inherits all history, meeting contributions, and intelligence from the loser. Merge is permanent and writes an audit event.
• Archive — Topics that are no longer active can be archived. Archived topics are excluded from future AI matching but their history is preserved. They can be unarchived if the theme re-emerges.

Best Practices#Copy link

Start with topics at workspace creation. Before your first recorded meeting, set up the 4–8 topics you expect to discuss most. This gives the AI meaningful anchors and produces cleaner automatic matches from day one.

Keep topic names specific but durable. “Website Redesign” is better than “Project” (too vague) and better than “Website Phase 2 Responsive Design Sprint” (too narrow for a recurring theme). Aim for names that will still be accurate five meetings from now.

Review proposed topic matches early. During the post-meeting review workflow, confirming or redirecting proposed topic links only takes a moment but has compounding value — each confirmed link improves the Topic Brief and the workspace’s AI context going forward.

Merge duplicates as soon as you spot them. When two topics diverge under slightly different names, merge them promptly. A fragmented topic registry means the AI has to work harder and the Topic Brief for each topic is thinner than it should be.

Use topic briefs before calls. Before a recurring meeting with a client, pull up the relevant topic briefs in Knowledge → Topics. The briefs surface what was decided, what’s still open, and what commitments were made — giving you a fast, accurate pre-call briefing.

Let topics grow organically after the kickoff set. You don’t need to pre-define every possible topic. When the AI encounters a theme that doesn’t match anything, it will propose a new topic. Review and accept these proposals as the workspace evolves — your topic registry should reflect how the engagement actually unfolds, not just how you expected it to.

Related Articles#Copy link

• Post Meeting Process — How to review and confirm extracted topics and intelligence items after a meeting
• Understanding Meeting Intelligence Capture — The full end-to-end journey from calendar to workspace knowledge
• Organizing Workspace Topics — How to manage the canonical topic registry in your workspace
• Workspace Context & Intelligence — How topics connect to the four types of workspace knowledge