> ## Documentation Index > Fetch the complete documentation index at: https://docs.callers.ai/llms.txt > Use this file to discover all available pages before exploring further. # Path to production > A go-live checklist for Callers agents: numbers, testing, voicemail, compliance, scheduling, integrations, publishing, and monitoring. Building an agent that works on a test call is the first step. Running it reliably on thousands of real conversations takes a few more. Use this checklist before you start a new agent or scale an existing one.
    From first draft to full volume ## Before you launch **1. Set up phone numbers and caller ID** > [!TIP] > > The agent has at least one phone number assigned (or a SIP trunk for your own carrier). > [!TIP] > > For outbound agents with several numbers, a dialing strategy is chosen: **Rotate Numbers**, **Prefer Local Numbers**, or **Auto Optimize**. > [!TIP] > > Numbers are checked for spam flags. Replace any number marked **Spam Risk Detected**. > [!TIP] > > For inbound agents, the number routes to this agent and **Max incoming calls in parallel** fits your expected volume. See [Phone numbers and caller ID](https://docs.callers.ai/documentation/contacts/phone-numbers-and-caller-id.md) and [Improve answer rates](https://docs.callers.ai/documentation/best-practices/improve-answer-rates.md). **2. Test across scenarios** > [!TIP] > > Test calls cover the main path to the goal from start to finish. > [!TIP] > > Each scenario is triggered at least once: common questions, objections, "not interested," "call me later," wrong number, and requests to stop calling. > [!TIP] > > You've tested interruptions, silence, and off-script answers. > [!TIP] > > Test variables use realistic values, including empty optional fields. > [!TIP] > > You've tested in each language the agent will speak. > [!TIP] > > Transfers reach the right number or queue, with the right caller ID. See [Testing your agent](https://docs.callers.ai/documentation/build-your-agent/testing-your-agent.md) and [Prompting for voice](https://docs.callers.ai/documentation/best-practices/prompting-voice-agents.md). **3. Decide voicemail behavior** > [!TIP] > > **Leave message if reaching voicemail** is on or off by choice, not by default. > [!TIP] > > If on, the strategy is set: **Whenever voicemail is detected**, **First detected voicemail only**, or **On the final scheduled call (if voicemail is detected)**. > [!TIP] > > The voicemail message is short, includes who is calling and why, and doesn't include sensitive information. See [Voicemail and screening](https://docs.callers.ai/documentation/voice/voicemail-and-screening.md). **4. Confirm compliance settings** > [!TIP] > > You have the consent required to contact everyone on your list, on every channel you use. > [!TIP] > > **Enable Account Do Not Call (DNC) List** and **Update Account Do Not Call (DNC) List Automatically** are on. > [!TIP] > > For US contacts, **Check US Federal Do Not Call (DNC) Registry** (and **Include State Do Not Call (DNC) Lists** if needed) are set according to your obligations. > [!TIP] > > Recording settings and any recording notice match your legal requirements. > [!TIP] > > Sensitive data detection is on if callers may share card numbers or PINs. > [!TIP] > > Guardrails cover topics the agent must not discuss. See [Compliance](https://docs.callers.ai/documentation/best-practices/compliance.md). **5. Set availability and retries** > [!TIP] > > Calling hours match your business and legal limits. Use **Use Contact's Time Zone** for contacts in several time zones. > [!TIP] > > **Total Dial Attempts**, **Min Wait Time Between Dial Attempts**, and **Max Daily Dial Attempts** are set on purpose (defaults: 5 attempts, 90 minutes, 3 per day). > [!TIP] > > You've decided whether to continue dialing after a contact asks to be called later. See [Scheduling and retries](https://docs.callers.ai/documentation/contacts/scheduling-and-retries.md). **6. Set concurrency** > [!TIP] > > **Limit Concurrent Calls** and **Max Outgoing Calls in Parallel** match what your team, transfer targets, and APIs can handle. > [!TIP] > > If the agent transfers to people, there are enough people available for the call volume. **7. Define success and post-call analysis** > [!TIP] > > Success criteria describe what "goal reached" means in concrete terms. > [!TIP] > > Custom outcomes (for example, "Follow-Up Scheduled") exist for each result you'll act on. > [!TIP] > > Key questions capture the data you need downstream, with the right answer types. > [!TIP] > > You've reviewed the outcome and summary of your test calls and they're correct. See [Success criteria](https://docs.callers.ai/documentation/results/success-criteria.md) and [Post-call analysis](https://docs.callers.ai/documentation/using-callers/key-questions.md). **8. Test webhooks and integrations** > [!TIP] > > Each in-call action has been tested with the **Test** button and on a live test call, including error and timeout cases. > [!TIP] > > Post-call actions run for the right outcomes and write to the right place in your CRM or spreadsheet. > [!TIP] > > Webhooks are subscribed to the events you need and your endpoint returns success quickly. > [!TIP] > > Your endpoint validates webhook signatures. > [!TIP] > > Test call data is filtered out downstream (payloads include `isTestCall`). See [Webhooks](https://docs.callers.ai/documentation/webhook/webhook-overview.md) and [Validating signatures](https://docs.callers.ai/documentation/webhook/validating-webhook-signatures.md). **9. Publish the version** > [!TIP] > > The tested draft is published, so edits you make later don't reach live contacts until you publish again. > [!TIP] > > Anyone else editing the agent knows which version is live. If you don't see version controls in the agent header, contact [support@callers.ai](mailto:support@callers.ai). See [Versions and split traffic](https://docs.callers.ai/documentation/build-your-agent/versions-and-split-traffic.md). ## Start small, then scale Don't launch to your full list on day one. **1. Run a pilot** Start the agent with a small batch of real contacts, such as 50 to 200, during normal calling hours. **2. Review every pilot conversation** Read transcripts and listen to a sample of recordings. Check outcomes, summaries, and that data reached your systems. **3. Fix and republish** Rewrite steps where contacts hang up or get confused. Add scenarios for questions you didn't expect. Publish the new version. **4. Increase volume gradually** Add contacts in larger batches and raise concurrency step by step. Watch answer rates and spam flags as volume grows. ## Monitor after launch | What to watch | Where | Why | | - | - | - | | Connect rate and success rate | Agents list and agent results | Spot drops in answer rate or conversion early. | | Drop-offs, hang-ups, and interruptions per step | **Insights Mode** in the agent header | Find the steps to rewrite first. Insights reset when the script changes. | | AI suggestions | Agent header | Get suggested fixes for high drop-off, high interruption, and missed questions. | | Outcomes and trends | [Analytics](https://docs.callers.ai/documentation/using-callers/analytics.md) | Track results across agents and over time. | | Agent and number alerts | [Notifications](https://docs.callers.ai/documentation/account/notifications.md) | Get told when an agent stops, runs out of contacts, has missing variables, or a number is flagged as spam. | | Webhook and action logs | Webhook settings and call transcripts | Catch failing integrations. | > [!TIP] > > Use [Studio Chat](https://docs.callers.ai/documentation/build-your-agent/studio-chat.md) to ask questions about recent results, such as "Analyze recent calls" or "Find customer objections." ## Related > **[Compliance](https://docs.callers.ai/documentation/best-practices/compliance.md)** > > Consent, DNC, opt-outs, and recording. > **[Improve answer rates](https://docs.callers.ai/documentation/best-practices/improve-answer-rates.md)** > > Local presence, retries, and pre-call chat. > **[Testing your agent](https://docs.callers.ai/documentation/build-your-agent/testing-your-agent.md)** > > Test calls, web calls, chats, and emails. > **[Agent results and insights](https://docs.callers.ai/documentation/using-callers/agent-insights.md)** > > Results, filters, and insights mode. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.