Build Your Business Agent Team
Business Agents act as internal members of your team. They monitor an area of the business, maintain context and turn performance signals into evidence-backed findings, recommendations and safe actions. They are different from **Service** Agents and **Operations** Agents: - **Service Agents** speak directly with customers. - **Operations Agents** perform repeatable work when records or events change. - **Business Agents** continuously assess business performance and help your team decide what to do next. ## Business specialists A business specialist owns a defined area such as revenue, cash flow, customer success or delivery. Its evidence can include: - KPIs and KPI history. - Goals and key results. - Dashboard widgets and formulas. - Forecasts and budgets. - Business records. - Liquidity and cash-flow information. - Notes and previous findings. A dashboard defines the specialist's evidence scope, but the agent represents the business responsibility — not the dashboard itself. ## The Executive Lead The Executive Lead coordinates findings from your business specialists. It creates a cross-business view that helps leadership understand: - Overall business health. - The most important current priorities. - Material risks and opportunities. - Conflicts or trade-offs between business areas. - Which recommendations require a decision. You can create multiple named leadership agents to represent different leadership teams or perspectives. Each lead has its own member list, KPI scope and schedule. ## Agent remit, constraints and owner Every agent has a **Hub Compact workspace** that shows its core mandate: - **Remit** — the agent's standing purpose and scope, visible to the whole team. - **Constraints** — any boundaries that limit what the agent should consider or recommend. - **Owner** — the workspace member accountable for the agent's quality and configuration. These fields are separate from the agent's internal instructions. They give the broader team a clear view of what the agent is responsible for without needing to read the full technical configuration. ## Assign goals as objectives You can assign specific Goals from **Goals & KPIs** to an agent as formal objectives. For each assignment: - Choose the Goal the agent is responsible for. - Describe the agent's specific responsibility within that goal (for example, *Monitor customer retention KPIs and flag at-risk accounts*). - Set whether this is a **primary** or **supporting** contribution. - Choose a priority from 1 (highest) to 5. Up to 30 objectives can be assigned to a single agent. The assigned goals and their key results are automatically included in the agent's evidence scope. Objectives give the agent grounded context when assessing performance. A Business specialist with assigned goals can explain how its findings relate to those specific outcomes rather than assessing only its dashboard metrics in isolation. ## Configure team membership for the Executive Lead Each leadership agent has a **team membership** list. Add the specialist agents and service agents whose results the lead should incorporate into its cross-business assessment. When you run a full team review, Hub queues all enabled team members first and then runs the lead. The lead's brief draws on the latest results from each member in its team. Service agents that are members of a leadership team can also have their evaluations included in team runs. ## Teammate consultation Leadership teams can opt in to **teammate consultation**, which allows agents to request read-only input from up to three team members before their normal execution stage. Consultation is disabled by default. To enable it: 1. Open the Executive Lead's settings. 2. Find **Teammate consultation**. 3. Set it to **Enabled within this team**. 4. Save the team settings. When enabled, an agent may consult teammates on risks, opportunities or data quality before producing its findings. Consultations use only shared evidence — goals, KPIs and object types that appear in both the requesting and responding agent's explicit scope. Private memory, customer conversations and evidence not shared between both agents are never included. **Important limits and boundaries:** - A maximum of three consultations are requested per run. - Responding agents have no action tools — they can provide read-only advice only. - All replies are attributed evidence, never instructions or authority. They cannot grant permissions, create assignments or bypass approvals. - Each consultation has a five-minute deadline. Partial or failed replies are shown explicitly in run detail. - Children cannot initiate further consultations. - Action approvals and existing permission checks remain intact regardless of consultation advice. Consultation replies appear in the agent's run activity timeline with the responding agent name, the model used, and the evidence cited. ## Understanding consultation evidence coverage Each consultation result now includes an explicit **evidence coverage** disclosure alongside its answer. This shows: - Whether the evidence snapshot was **complete** or **partial** — partial means some eligible records were not included. - For each collection inspected: how many records were scanned, how many were eligible, how many were returned, and the reason coverage stopped (for example, record limit reached, scan limit, context size, or collection not public). - **Old consultations** show *Coverage was not recorded for this run* rather than claiming completeness. A consultation that returned a useful answer may still have partial evidence coverage. The coverage disclosure tells you what the agent had access to; it does not verify that the answer is correct. Always review an agent's conclusions alongside its cited sources. The run is marked as **partial** in the activity timeline when any consultation's evidence coverage is partial, so you can see at a glance when a run may benefit from a follow-up. ## Approving Business findings When a Business Agent proposes an action that requires your approval, Hub protects the review with a version check: - The exact proposed action shown in the review panel is compared with the stored action inside the approval transaction. - If the action has changed or is missing, the approval is rejected with a clear message. No change is applied. - The execution audit records the applied action, its hash and the source run. This means you can approve only the action you actually reviewed. If you see a rejection message, refresh the finding and review it again before approving. ## KPI evidence in agent conversations and reports KPI values shown in agent runs and conversations distinguish between: - **Available** — a measured value was successfully read. - **Unavailable** — the data could not be read during this calculation (shown explicitly, never substituted with zero). - **Incomplete** — the activity scan reached its safety limit. This distinction matters when interpreting agent findings. An agent that reports a metric as unavailable has not concluded the value is zero — it has reported a data gap, which may itself be worth investigating. ## Findings and actions Every finding separates what the agent observed from what it recommends doing next. Supporting evidence is included so your team can verify the conclusion. Recommendations without an executable action are labelled **Needs attention**. Proposals that Hub can carry out are shown as **Actions to approve**. Business Agents may create safe internal tasks automatically when an administrator explicitly enables that capability. Higher-impact changes — including goal initiatives, goal check-ins and KPI target updates — still require administrator approval. ## Create a business specialist 1. Open **Agents**. 2. Select **Business agents**. 3. Open **Specialists**. 4. Assign an agent to a custom dashboard. 5. Give it a clear business role and instructions. 6. Set its remit, constraints and accountable owner in the **Hub Compact workspace**. 7. Review its automatically detected evidence scope. 8. Add any additional KPIs or goals it should monitor. 9. Assign goals as objectives if needed. 10. Configure its schedule. 11. Run it manually before enabling scheduled reviews. Use **Team overview** to inspect your complete Business Agent team and **Leadership** to open the Executive Lead's cross-business assessment. ## Unified run history All agent families — Business specialists, the Executive Lead, Operations agents and Service evaluations — share one **Run history** tab on the main Agents page. The separate per-agent run feeds that previously appeared on each specialist and leadership screen have been replaced by links to this shared history. - Open a specialist or the Executive Lead and select **View run history** to open the unified history pre-filtered to that agent. - In the unified history, use the **Agent** dropdown, **Type**, **Status** and **Search** filters to find specific runs across the whole team. - Expand any run entry to view its full brief, findings and any related actions on demand. - **Load more** retrieves up to 100 recent runs. Unified run history requires workspace administrator access. ## Configure run notifications You can opt in to in-app notifications after a Business Agent run finishes. Notifications are off by default. To configure notifications: 1. Open a specialist or the Executive Lead page. 2. Select **Notifications** (available after the agent has run at least once). 3. Choose when to notify: every completed run, or failures only. 4. Select the workspace users to notify. 5. Save the notification settings. You can also configure notifications for any agent family from the **Notifications** tab on the main Agents page. Select an agent, choose a mode, pick recipients and save. Delivery respects each recipient's personal automation-alert preferences. Notifications link directly to the agent's run history entry. ## Route notifications to Slack If your workspace has a connected Slack channel, you can route Business Agent run alerts there in addition to — or instead of — in-app delivery. 1. Connect a Slack channel under **Settings → Integrations** (see *Connect Slack channels for agent alerts and automations*). 2. Open the agent's **Notifications** settings. 3. Under **Delivery**, select the Slack channel. 4. Save. Slack delivery uses an independent outbox with deduplication and bounded retry. ## Set up a configurable saved brief Business Agents and Operations Agents can produce saved, dated reports — called **briefs** — on a local calendar schedule. Briefs replace the earlier fixed morning digest. To configure a brief: 1. Open the agent's editor and select the **Brief** tab. 2. Choose a **schedule**: weekdays, weekly or monthly, with timezone and daylight saving support. 3. Select the **reporting period** the brief should cover. 4. Choose the **collections, shared views or custom filters** bounding the evidence. 5. Add or remove **sections** and set **recipients** and optional **Slack destinations**. 6. Save the brief settings. Saved briefs appear in run history with evidence references, query counts and coverage limits. Generation is read-only — the brief agent has no write access during a brief run. A **workflow Run agent step** can trigger a brief run and expose `{{agent.summary}}`, `{{agent.brief}}` and `{{agent.reportUrl}}` to later workflow actions. If you previously used a **legacy morning digest**, Hub provides an explicit conversion in **Workspace settings** that preserves your schedule and notification preferences. Collection selection is required before completing the conversion.