Playbooks

Playbooks are step-by-step guides for building Duvo agents around specific business outcomes. Each playbook covers a real workflow end to end: why to automate it, what you’ll build, how to set it up, and how to handle the edge cases that come up in practice. Use a playbook when you want to automate a recognizable business goal — not a feature, but an outcome: “monitor inventory and alert on low stock”, “process orders from a shared inbox”, “triage customer complaints and route them to the right team”.

---

​

How Playbooks Are Structured

Every playbook follows the same structure so you can quickly find what you need:

---

​

Playbook Categories

Playbooks are organized by the type of outcome they deliver.

---

​

Inbound Intake

Turn incoming emails, documents, and forms into structured records — without manual rekeying.

---

​

Documents

Process documents at scale — extract structured data, validate against master data, and route exceptions for human review.

---

​

Reporting

Pull data from your systems on a schedule and deliver it where it is needed — Slack summaries, email digests, spreadsheet exports. Coming soon: Forecasting and Alerting — monitor warehouse data and alert on threshold breaches.

---

​

Operations

Handle the exception-heavy workflows that fall through the cracks: stuck transactions, failed items, aging cases, supplier follow-ups, QA spot checks. Coming soon: Exception Handling, Supplier Follow-up, QA Automation.

---

​

Customer-Facing

Route, triage, and respond to inbound customer requests consistently and at scale.

---

​

Per-Connection Outcome Examples

Short outcome-based examples for the highest-usage connections. For step-by-step tutorials, follow the links to the full workflow guides.

---

​

Slack

These workflows use Slack as an operational hub — team members trigger agents, receive results, and approve actions without leaving Slack. On-demand reports from shared channels A team member mentions the Duvo app with a question (“how many orders are stuck in pending?”) in any enabled channel. The agent queries your data source and posts a formatted answer back in the thread within minutes. No dashboards. No analyst required. See Slack Mention Workflows. Request intake and routing Requests that arrive in #helpdesk, #ops-requests, or similar shared channels are read, classified, and routed to the right person or ticketing system automatically. The requester gets an acknowledgment in the thread; the right team gets a ticket or DM. See Slack Mention Workflows — Request Intake and Routing. Automated status updates on a schedule Every Monday morning, a structured summary of last week’s key metrics — revenue, exceptions, open cases — is posted to your team’s Slack channel. Built from live warehouse or CRM data; no manual preparation needed. See Snowflake Workflows — Scheduled KPI Report. Threshold alerts When inventory falls below the reorder point, a fraud rate crosses a limit, or an error count spikes, your team gets a Slack notification immediately rather than discovering the problem in a weekly report. See Snowflake Workflows — Threshold Alert.

---

​

Gmail

These workflows turn your inbox from a queue of interruptions into a structured input channel for your business systems. Inbox triage Incoming emails are classified by category (support request, lead inquiry, invoice, newsletter), labeled in Gmail, and routed automatically — forwarded to the right team, logged to a spreadsheet, or saved to Drive. You open your inbox to find it already organized. See Gmail Workflows — Inbox Triage. Draft replies to common inquiries For routine customer questions, the agent drafts a reply using your product documentation and preferred tone, saves it to Drafts, and waits for you to review. Complaints are flagged in Slack. Nothing is sent until you approve. See Gmail Workflows — Drafting Replies. Email order intake Order emails arriving in a shared inbox are read, structured fields extracted, records written to your order management system, and anomalies escalated for human review. The sender receives a confirmation or a clarification request automatically. See Email Order Intake.

---

​

Snowflake

These workflows read directly from your data warehouse to power reports, alerts, and exports — without requiring a data engineer or a BI dashboard. Scheduled KPI report Every Monday morning, a weekly summary of revenue, order count, exception rate, and week-over-week changes is posted to your team’s Slack channel or emailed to stakeholders. The agent queries Snowflake, formats the results, and distributes them automatically. See Snowflake Workflows — Scheduled KPI Report. Inventory threshold alerts When SKU inventory drops below the reorder point, the agent posts a Slack alert with the specific items and quantities. Alerts are deduplicated so the same issue does not notify repeatedly before anyone can act. See Snowflake Workflows — Threshold Alert. On-demand data exports to Google Sheets A stakeholder needs a fresh data pull — customer list, transaction export, regional breakdown. The agent runs the query and writes results into a named Google Sheet, including a “last refreshed” timestamp. Run on demand or on a schedule. See Snowflake Workflows — Export to Google Sheets.

---

​

Scheduling Patterns

Most automations follow one of these scheduling patterns. Choose the one that fits your workflow’s cadence. Recurring digest (daily or weekly) Run an agent on a fixed schedule — every Monday at 7am, every weekday at 8am — to pull data, format a summary, and distribute it. Ideal for KPI reports, inbox roundups, and standing briefings that would otherwise require manual preparation. See Scheduling Agents for setup instructions. Continuous monitoring with threshold alerting Run an agent every hour (or every 15 minutes for time-critical metrics) to poll a data source and act only when a condition is met. The agent checks silently and does nothing if everything is within bounds, so notifications are signal not noise. See Scheduling Agents for setup, and Snowflake Workflows for an example using warehouse data. Event-driven trigger with scheduled follow-up An agent starts when an event occurs (a new email, a Slack mention, a file upload) and then self-schedules a follow-up check. For example: process an order email immediately, then check 48 hours later whether a clarification response arrived. No separate cron job needed. See Event-Driven Triggers and Scheduling Agents — Self-Scheduling. Batch processing window Run a large-volume workflow once per day during off-peak hours — process all unread order emails overnight, reconcile records at 2am, generate invoices for all completed runs at end of day. Keeps interactive workloads unaffected while ensuring nothing accumulates indefinitely. See Scheduling Agents for setup instructions.