Asking Effective Questions

Agency Hero’s AI has access to far more than you might expect: months of meeting transcripts, every extracted decision and risk, recurring topic patterns, task lists, deal pipeline data, uploaded documents, memos, and your full org knowledge base. The quality of your results depends on how well your question matches what the AI can actually do.

This guide covers two things: what kinds of questions unlock the AI’s full capability, and the difference between in-workspace chat and cross-workspace chat — because where you ask from changes what the AI draws on.

New to Chat? Start with Getting Started with Chat first. To understand how messages get routed to the right capability, see How Skill Routing Works.

Two chat modes — and why they matter#Copy link

The single most important thing to understand about Agency Hero chat is that where you’re chatting from changes the AI’s context. There are two distinct modes.

In-workspace chat#Copy link

When you open chat from inside a specific workspace — say, Acme Corp — the AI’s context is automatically loaded with everything that workspace holds:

• The workspace’s full meeting history and transcripts
• All extracted intelligence: decisions, risks, open questions, commitments, objections, assumptions, blockers, milestones, and action items
• Recurring topics and the detailed briefs built up across all meetings
• The workspace’s tasks (with statuses, owners, and due dates)
• Stakeholders and org structure for that client
• Uploaded knowledge artifacts: SOWs, proposals, commercials, objectives, playbooks
• Memos and research notes saved in that workspace
• Deal pipeline stage and associated signals

You don’t need to say “in Acme Corp” with every question — the workspace is already the AI’s frame of reference. This makes in-workspace chat ideal for deep, client-specific questions.

Example questions you can ask from inside the Acme Corp workspace:

“What commitments have we made to this client that we haven’t yet delivered on?”

“Walk me through every decision we’ve made about their API migration — oldest to newest.”

“What objections has Acme raised across all our sales meetings, and how did we respond each time?”

“Which risks are still open from our last three meetings?”

“Who are the key stakeholders in this account, and what topics do they tend to raise?”

“What does the SOW say about scope for Phase 2? Does that match what we’ve agreed to in meetings?”

“Give me a timeline of every milestone we’ve committed to and flag any that are overdue.”

“What’s this deal’s current stage, and what open blockers could stall it?”

“Summarize the last 90 days with Acme — key decisions, risks, and outstanding action items.”

Cross-workspace chat#Copy link

When you chat from the home or global context — not from inside any specific workspace — the AI can draw from all workspaces you have access to. This is the right mode for portfolio-level questions, cross-client comparisons, and personal task management across your entire book of work.

You can also explicitly name a workspace in any question to scope it precisely, even from a global context.

Example questions that work across workspaces:

“Which of my active workspaces has the most unresolved risks right now?”

“Show me all overdue tasks assigned to me across every workspace.”

“Which clients have had the fewest meetings in the last 60 days? I want to identify accounts that might be going quiet.”

“Across all my workspaces, what commitments are outstanding this week?”

“Compare the open blockers in the Acme Corp and Vertice Labs workspaces.”

“Which workspaces are in the proposal or negotiation stage right now?”

“Give me a status summary of all my active client engagements — one paragraph each.”

“Do I have any meetings coming up this week across all workspaces?”

“Search all workspaces for any decisions made about pricing in the last 90 days.”

“Which clients have raised security or compliance concerns in meetings?”

The cross-workspace rule: By default, if you’re in a workspace, the AI stays scoped to it. If you want broader results from the home context or want to override scope, use phrases like “across all my workspaces”, “everywhere”, or name the specific workspace you mean.

What the AI can actually access (and how to ask for it)#Copy link

Understanding what’s in the AI’s reach helps you ask questions that get genuinely impressive answers. Here’s a map of what the AI knows and how to unlock it.

Meeting history and transcripts#Copy link

The AI has access to full meeting transcripts and executive summaries — not just recent ones, but months of history. You can ask it to reason across time.

💡 Pro tip: If you’re already viewing a specific meeting in Agency Hero, the AI has that meeting’s full transcript, summary, and intelligence items loaded. You can ask “What action items came out of this meeting?” without mentioning the meeting at all.

Intelligence items#Copy link

Every meeting is processed to extract nine types of intelligence: decisions, risks, questions, commitments, objections, assumptions, blockers, milestones, and action items. These are stored as individual, searchable items — and the AI can query them specifically.

Topics#Copy link

Topics are recurring themes that the AI maps across meetings over time — not one-off mentions, but persistent threads that keep surfacing. By the time you’ve had ten meetings on a client, the AI has a rich, evidence-backed brief on each topic.

“What does this workspace’s topic history show about how our relationship with this client has evolved?”

“Which topics have come up in every single meeting with this client?”

“Tell me everything we know about the ‘API performance’ topic — what’s been said, decided, and flagged as a risk.”

“What new topics have emerged in the last 30 days that weren’t discussed before?”

Tasks#Copy link

The AI knows about every task in a workspace: its title, description, status, assignee, due date, and which meeting or intelligence item it came from.

“What tasks are overdue in this workspace and who owns them?”

“Show me all tasks assigned to Jordan that don’t have a due date yet.”

“Which tasks came out of last Tuesday’s planning meeting that haven’t been started?”

“Are there any high-priority tasks that have been open for more than two weeks?”

Knowledge artifacts (SOWs, proposals, commercials, objectives)#Copy link

Uploaded documents and structured knowledge artifacts are fully searchable by the AI — including SOWs, proposals, commercial terms, objective frameworks, and org templates.

“What does the SOW say about our deliverables for Phase 3?”

“Does the proposal include anything about SLA guarantees? What exactly does it say?”

“Compare what we promised in the proposal to the decisions we’ve actually made in meetings.”

“What are the stated objectives for this engagement, and which ones don’t yet have supporting tasks?”

People and org structure#Copy link

“Who are the key stakeholders in this workspace and what roles do they hold?”

“Which people from the client side have been most active in meetings?”

“Have we ever spoken directly with their CTO, or only with the project manager?”

Memos and research notes#Copy link

Memos saved in a workspace — whether written manually or generated from meeting outputs — are available for the AI to retrieve and reason over.

“Are there any research memos saved about the competitor landscape for this client?”

“Show me any notes we’ve saved about the client’s internal buying process.”

“Find the memo about our go-live strategy and summarize the key points.”

Making questions more precise#Copy link

Name the type of thing you want#Copy link

Agency Hero stores information in distinct places. Naming the type of data you want sends the AI straight to the right source — no guessing.

Add time ranges for historical queries#Copy link

The AI always knows today’s date and your timezone, so relative terms work fine. For historical queries, a specific range sharpens the results.

Be explicit about actions vs. lookups#Copy link

The AI both reads (search, summarize, retrieve) and writes (create tasks, capture decisions, draft memos). Mixing them in one message can blur intent.

For any action that changes data, the AI will show a confirmation card before saving. You can review and edit before anything is committed.

Ask one focused question at a time#Copy link

Multi-part questions can lead to partial answers. If you need multiple things, number them explicitly.

Context the AI loads automatically#Copy link

You don’t need to explain the basics every time. The AI already knows:

• Your identity — your name, timezone, and which workspaces you can access
• Current workspace — when you’re in a workspace, all questions are automatically scoped to it
• Current page — if you’re viewing a meeting, that meeting’s full intelligence is already loaded; if you’re on a task, the task details are in context
• Today’s date — relative terms like “this week” or “last month” work without explanation

You only need to add context when you want something outside your current view, or when you want to override the default scope.

Quick reference: question patterns that work well#Copy link

In-workspace (deep client / project focus)#Copy link

Decisions and intelligence

• “What decisions have we made about [topic] in this workspace?”
• “What risks are currently open and who flagged them?”
• “What commitments have we made that aren’t reflected in any tasks yet?”
• “What assumptions are we operating under that have never been validated?”

Meeting history

• “Summarize the last 90 days with this client — key decisions, risks, and action items.”
• “What changed in our discussions about [topic] from Q1 to Q2?”
• “Which meetings covered the contract renewal, and what was resolved vs. still open?”

Tasks and ownership

• “What tasks are overdue in this workspace?”
• “Show me all open tasks assigned to [person] with no due date.”
• “Which action items from the last three meetings haven’t been converted to tasks yet?”

Knowledge artifacts

• “What does the SOW say about [specific deliverable]?”
• “Does our proposal address their concern about [topic]? What does it say exactly?”
• “What are the stated objectives for this engagement?”

Deal and pipeline

• “What stage is this deal in, and what are the open blockers to moving forward?”
• “What objections has the client raised in the last three meetings?”
• “Prepare a quick briefing for a call with this client tomorrow.”

Cross-workspace (portfolio and personal management)#Copy link

Portfolio overview

• “Give me a one-paragraph status update on each of my active workspaces.”
• “Which workspaces have had no meetings in the last 30 days?”
• “Which of my clients are currently in a proposal or negotiation stage?”

Cross-client patterns

• “Have any of my clients raised concerns about pricing in the last quarter?”
• “Which workspaces have open risks that haven’t been addressed in 2+ weeks?”
• “Across all my workspaces, what topics keep coming up that I should pay attention to?”

Personal task management

• “What tasks are assigned to me across all workspaces, sorted by due date?”
• “Do I have overdue action items anywhere?”
• “What commitments do I personally own that are still outstanding?”

Meetings and scheduling

• “What meetings do I have coming up this week across all workspaces?”
• “Which client haven’t I met with in the longest time?”

When a response misses the mark#Copy link

If the AI returns results from the wrong scope, gives a partial answer, or asks for clarification:

1. Name the workspace explicitly — “I meant the Acme Corp workspace, not the current one.”
2. Narrow the type — “Just decisions, not tasks or risks.”
3. Add a time range — “Only from the past two weeks.”
4. Clarify the action — “I want to see it, not create it.”
5. Specify cross-workspace explicitly — “Search across all my workspaces for this.”

The AI maintains conversation history within a session, so you can refine in follow-up messages without starting over.

Next steps#Copy link

Once you’re comfortable asking good questions, the next article covers Searching and Retrieving Information — a deeper look at how search works across topics, documents, meeting summaries, and the intelligence database.