Understanding Intelligence Items

Every meeting is full of signal: commitments made, risks flagged, questions left open, competitors named, assumptions that should be validated. Agency Hero automatically surfaces all of this as intelligence items — structured, searchable records that persist across meetings and build a longitudinal knowledge base for your workspace.

This article explains what intelligence items are, how they are extracted, where to find them, and how to manage them.

What Are Intelligence Items?#Copy link

Intelligence items are atomic, structured facts extracted from meeting transcripts. Each item captures one specific claim, event, or observation — the kind of thing a careful analyst would write down after reviewing a recording.

Unlike a meeting summary (which is narrative prose), intelligence items are:

• Typed — every item has a category (decision, risk, objection, etc.) that shapes how it is stored, displayed, and used
• Structured — each type has specific fields (severity, ownership, status, due dates) rather than free text
• Trackable — items carry a lifecycle status so you can see what’s open, resolved, or dismissed
• Cumulative — confirmed items live in your workspace’s intelligence ledger and persist across every meeting

Over time, the ledger becomes a searchable record of everything important that has been discussed, decided, or flagged in your workspace.

Core distinction: Topics are the chapters of your workspace story (organizing themes like “Budget” or “Technical Integration”). Intelligence items are the sentences — the specific claims and facts that belong inside each chapter.

Intelligence Item Types#Copy link

The types available in your workspace depend on the workspace type (deal, project, ops, etc.) and your organization’s configuration. Here is the complete catalog.

Core Types — Available in All Workspaces#Copy link

These types are extracted for every workspace type.

Decision

A choice or resolution that was made during the meeting. A decision is only extracted if it was clearly agreed — proposed ideas that were not finalized are not captured as decisions.

• Decision type: scope, pricing, timeline, process, technology, resourcing, go/no-go, or other
• Status: proposed → decided → reversed / deprecated
• Example: “Agreed to push the launch to March due to team availability”

Risk

Something that could prevent the delivery of value that has been committed to. Agency Hero applies a strict “threatens paid delivery” test: vague concerns and sentiment signals are excluded. Only concrete threats to scope, timeline, quality, adoption, access, or payment qualify.

• Severity: likelihood (0–1) × impact (1–5 scale)
• Status: open → watching → mitigated / accepted / closed
• Example: “API documentation delay threatens the integration deadline”

Question

An open item that was explicitly unanswered at the end of the meeting and blocks something (a task, a decision, or a topic). Questions that were answered during the meeting are captured as decisions instead. “Task in disguise” requests (“Can you send me…”) are excluded — these become action items instead.

• Priority: P0 (urgent/blocking), P1 (important), P2 (nice to know)
• Status: unassigned → researching → answered / obsolete
• Example: “What is the data residency requirement?”

Proof

Evidence or validation of capability that was requested or offered — such as a case study, testimonial, metric, certification, or reference.

• Permission status: not asked → asked → approved / restricted / denied
• Example: “Client asked for a reference from a company in the healthcare space”

Expansion

An opportunity to expand the engagement, such as an upsell, cross-sell, renewal, or referral signal.

• Expansion type: upsell, cross-sell, renewal, referral
• Status: identified → qualified → proposed → won / lost
• Example: “They mentioned interest in adding the analytics module next quarter”

Deal-Specific Types — Deal Workspaces#Copy link

These types are available in deal workspaces and are optimized for tracking buyer dynamics and sales qualification.

Commitment

Something the prospect or seller explicitly agreed to do. Tracked by who made the commitment, the due date, and whether it is conditional on something else.

• Made by: prospect or seller
• Status: open → fulfilled / broken / superseded
• Examples:
  • “We’ll send the security questionnaire by end of week” (seller commitment)
  • “I’ll set up a meeting with the CFO” (prospect commitment)
  • “If the demo goes well, we’ll move to contract review” (conditional)

Objection

A concern, pushback, or hesitation that could block the deal. Classified by type and severity so you can prioritize your response strategy.

• Objection type: price, timing, authority, need, trust, competition, status quo, other
• Severity: low, medium, high, critical
• Status: raised → addressed / unresolved / deal blocker
• Examples:
  • “That’s more than we budgeted” (price, medium)
  • “We need this live by Q2 — that timeline is too long” (timing, high)

Assumption

A statement believed to be true that should be validated before the deal progresses. Assumptions can be stated (the prospect said it directly) or inferred (the AI detected an implicit belief).

• Source: stated or inferred
• Validated: yes / no
• Examples:
  • “They’re using Salesforce” (stated)
  • “They probably have budget authority” (inferred)

Dependency

Something that must happen or be true for the deal to progress. Classified by who owns the dependency.

• Dependency type: approval, technical, legal, budget, timeline, other
• Owner: prospect, seller, third party, unknown
• Status: pending → in progress → resolved / blocked
• Examples:
  • “CFO approval required for deals over $25k”
  • “IT must validate the integration approach”

Constraint

A hard or soft limitation on the engagement. Hard constraints cannot be negotiated; soft constraints indicate a preference.

• Constraint type: budget, timeline, technical, legal, resource, other
• Is hard: true / false
• Examples:
  • “Budget cannot exceed $50k” (hard)
  • “Prefer not to use AWS, but could if necessary” (soft)

Competitor

An alternative solution or competitor being evaluated. Includes direct competitors, indirect alternatives, the status quo (do nothing), and build-in-house options.

• Competitor type: direct, indirect, status quo, build in house
• Threat level: low, medium, high
• Examples:
  • “They’re also looking at Gong” (direct, high threat)
  • “They could just use spreadsheets” (status quo, medium threat)

Proof Point

A positive buying signal or evidence that your solution resonates with the prospect. Different from a Proof (which is evidence you deliver); a Proof Point is a signal from the buyer.

• Proof type: case study resonance, feature validation, ROI recognition, champion signal, other
• Signal strength: weak, moderate, strong
• Examples:
  • “That case study is exactly our situation” (case study resonance, strong)
  • “I’m going to push hard for this internally” (champion signal, strong)

Gap (Agent-Created)

A gap is different from every other type: it is not extracted from the transcript. Instead, after extraction, the AI agent analyzes what intelligence is present versus what is expected for the current deal stage and creates gap records for anything missing.

• Severity: critical, important, nice to have
• Status: open → addressed / dismissed / not applicable
• Auto-resolution: when you confirm an intelligence item of the missing type, the matching gap is automatically marked as addressed
• Example: “No budget assumptions captured — critical gap for Discovery stage”

Project-Specific Types — Project & Retainer Workspaces#Copy link

How Intelligence Items Are Extracted#Copy link

Extraction happens automatically as part of the post-meeting workflow that runs after every meeting is processed.

The Extraction Pipeline#Copy link

1. Transcript ready — Agency Hero receives the finalized meeting transcript
2. Topics extracted — The workflow first identifies the key discussion topics and assigns each a rank
3. Intelligence extracted — A specialized AI agent (the Intelligence Extractor) analyzes the full transcript and extracts items for each enabled type
4. Quality gates applied — Each candidate item must pass a five-part actionability filter before it is saved:
   • Has a future impact (changes what someone will do after the meeting)
   • Was not fully resolved during the meeting itself
   • Is not purely informational
   • Is actionable from the item text alone
   • Is grounded in what was actually discussed
5. Topic assignment — Each extracted item is assigned to the most relevant meeting topic
6. Confidence scored — Items are scored 0–1 based on how clearly the signal appears in the transcript:
   • 0.9+ Explicit, unambiguous statement
   • 0.7–0.89 Clear signal with minor interpretation needed
   • 0.5–0.69 Reasonable inference with some ambiguity
   • Below 0.5: not extracted
7. Proposed status — All extracted items start in a proposed state, waiting for your review

What Types Are Extracted?#Copy link

The types extracted depend on your workspace type and your organization’s configuration. Deal workspaces extract the full sales-focused taxonomy; project workspaces extract decisions, risks, questions, and project-specific types; ops workspaces extract their own set. Your admin can adjust which types are enabled per workspace type.

Quality Over Quantity#Copy link

Agency Hero’s extraction is intentionally selective. The AI applies strict rules specific to each type:

• Risks must pass a “threatens paid delivery” test — not every concern qualifies
• Questions must be genuinely unanswered at meeting end and gate something real
• Decisions must be clearly agreed, not just floated

This means you may see fewer items than you’d expect, but the ones you see are high signal.

Where to Find Intelligence Items#Copy link

Meeting Detail Page#Copy link

After a meeting is processed, the intelligence review section appears on the meeting detail page. Items are grouped into three confidence tiers:

• 🟢 High confidence (≥ 85%) — Items the AI is very confident about; bulk confirm available
• 🟡 Medium confidence (60–85%) — Items worth reviewing individually before confirming
• 🔴 Low confidence (< 60%) — Items that need attention; may need editing or dismissal

Each item shows its type badge, confidence percentage, summary text, and key payload details (e.g., severity for risks, who made a commitment, due dates).

Workspace Intelligence Ledger#Copy link

Once items are confirmed, they enter the workspace intelligence ledger — a persistent, cross-meeting record of everything important that has been captured for the workspace. You can browse the full ledger from your workspace’s intelligence page, filter by type or status, and search across all confirmed items.

Topic Pages#Copy link

Confirmed intelligence items are linked to workspace topics. When you open a topic page, you’ll see all intelligence items associated with that topic, regardless of which meeting they came from. This gives you a longitudinal view — for example, every risk or objection ever raised about “Pricing” across all your client conversations.

HubSpot Extension (Deal Workspaces)#Copy link

For deal workspaces connected to HubSpot, intelligence items also appear in the HubSpot UI extension panel on deal and contact records, giving your sales team access to meeting intelligence directly inside their CRM.

Reviewing, Confirming, and Dismissing Items#Copy link

Every extracted item starts as proposed — it needs a human review before it enters the canonical ledger. This Human-in-the-Loop (HITL) step keeps your workspace knowledge accurate.

Your Review Options#Copy link

Bulk Confirm#Copy link

For high-confidence items (≥ 85%), a Confirm All button lets you accept the entire group at once. This is safe to use when the items are clearly accurate — it saves time after meetings with many unambiguous signals.

Practical Tips#Copy link

💡 Review soon after the meeting. Intelligence items are most useful when they’re fresh. Try to review within 24 hours.

💡 Use Edit liberally. The AI’s phrasing may be close but not perfect. A quick edit to tighten the summary or correct a field is better than dismissing a valid item.

💡 Dismiss noise, not uncertainty. If an item is uncertain but real, edit it to add clarity. Dismiss only items that are wrong or irrelevant.

💡 Check low-confidence items carefully. These often surface genuine signals the AI is unsure about — they’re worth a look before dismissing.

How Intelligence Items Connect to Topics#Copy link

Topics and intelligence items work together to organize your workspace knowledge.

• During extraction, each intelligence item is assigned to a meeting topic (the primary discussion thread it belongs to)
• When confirmed, the item is linked to the corresponding workspace topic — a persistent, canonical theme that spans multiple meetings
• A confirmed item can be linked to multiple workspace topics if it is relevant to more than one (e.g., a risk about the integration could belong to both the “Technical” and “Timeline” topics)

This means browsing a workspace topic surfaces not just one meeting’s content, but the full history: every decision, risk, question, or objection that has ever been raised about that theme across all meetings in the workspace.

How Intelligence Accumulates Over Time#Copy link

The intelligence ledger is designed to grow with your workspace. Each confirmed item persists indefinitely, building a structured record that improves over time:

• Decisions build a complete log of what was agreed and why
• Risks show how concerns evolved — from open to mitigated to closed — across multiple meetings
• Questions track open items from their first mention to eventual resolution
• Commitments (deal workspaces) create an auditable record of what was promised and by whom
• Gaps (deal workspaces) highlight where your deal knowledge is thin, prompting targeted questions in future meetings

Over many meetings, this accumulated intelligence becomes the foundation for derived features like topic briefs (AI-generated summaries of what’s been discussed and decided about a topic) and deal briefs (a synthesized overview of a deal’s status drawn from confirmed intelligence).

💡 The more you confirm, the more powerful your workspace becomes. Each confirmed item enriches topic briefs, feeds AI-assisted analysis, and improves the context available to every downstream feature.

Workspace Type Availability Summary#Copy link

Your organization’s configuration may enable or disable specific types per workspace type.

Related Articles#Copy link

• Reviewing Your Meeting Summary — How to navigate the full post-meeting workflow
• Understanding Workspace Topics — How topics organize intelligence across meetings
• Using the Intelligence Ledger — Browsing, filtering, and searching confirmed intelligence
• Deal Briefs — How confirmed deal intelligence is synthesized into a deal overview