> ## Documentation Index > Fetch the complete documentation index at: https://docs.tranthor.com/llms.txt > Use this file to discover all available pages before exploring further. # Engagement Team > Set a business goal and let your AI engagement team plan campaigns, draft the messages, and propose them for your approval. The engagement team is a group of AI agents that work toward a business goal you set. They study your customers, decide what to send and to whom, and prepare campaigns. By default they stop there: **nothing is sent until you approve it.** Open it from **Agents** in the sidebar. ## How it works On your first visit, the Agents page shows the four steps: **Team goal**, **Objectives**, **Campaign drafts**, and **Reviewed & approved execution**. 1. **Team goal.** You adopt one business goal for the workspace. 2. **Objectives.** Each objective focuses on one customer outcome, such as winning back lapsed customers. 3. **Campaign drafts.** On a schedule, each objective assesses your customers and produces a proposal. 4. **Reviewed & approved execution.** You review each proposal. Approved campaigns are sent within your [operating limits](/guide/contact-controls). ## Set up your team If you have not set up a team yet, Agents opens a guided chat titled **Build your customer-engagement team**. Tell it your business goals, and it will agree on objectives, messages, and operating limits with you. You can ask questions about the limits at any point, and answer its clarifying questions in your own words. When the plan is ready you see **Review Your Engagement Team Plan**. It shows your goal, operating limits, objectives, and message drafts. **Nothing changes until you approve the plan.** After you approve, the page confirms **Your Engagement Team Is Configured**. ### Your goal A goal has a title, an optional description of the strategy, and a time horizon of 7 to 365 days (the default is 60). You can also pick a target metric, with a baseline and target value: * 30-day or 90-day repeat purchase rate * Average order value * 30-day churn rate reduction * Revenue per recipient * Net margin retention * A custom metric of your own ## Follow your objectives The Agents board groups objectives into lanes. Lanes with nothing in them are hidden. | Lane | What it means | | - | - | | **Ready to set up** | Available to set up, but not configured yet | | **Active Objectives** | Set up and scheduled | | **Evaluating / Running** | An assessment is running now | | **Needs Review** | A person needs to act, for example on a pending approval or a blocking issue | Select an objective to open its details. The tabs are **Activity** (every assessment cycle), **Proposals**, **Evidence & Outcomes**, and **Goal Direction**. * **Assess now** starts an assessment cycle immediately. * **Pause Objective** stops scheduled cycles and new candidates. Everything recorded so far is kept. You can add a reason, and resume at any time. Every cycle ends in one of three ways: it **acts** (creates a proposal), **waits** (takes no action this cycle), or **escalates** to you. ## Review proposals A proposal is a campaign the team wants to send, with its audience and message. Pending proposals appear on the objective's **Proposals** tab and on [Home](/guide/dashboard-home-ai-copilot). The bell in the header lists items that are waiting for you. For a campaign proposal you can: * **Approve Proposal** to send it within your limits. * **Request Changes** with a note on what to change. * **Reject** with an optional reason. **Experiment proposals cannot be approved.** They exist so you can review them and give feedback. They are never launched automatically. Proposals expire. If one has expired, or your goal, limits, or objective changed while you were reviewing, approval is declined and the draft is kept. Select **Re-evaluate Proposal** to get a fresh proposal under your current settings. ## Review mode and autonomous sending Every objective has a **Sending Mode**: * **Propose Only (Requires Review)**: the team drafts, you approve. This is the default, and it shows as **Manual Review Only (Drafts)** on the operating limits card. * **Autonomous Send (Under Policy)**: eligible campaigns send without per-campaign approval. We recommend running in review mode first, so you can judge the quality of proposals before you let the team send on its own. ### When an objective can send on its own An objective sends automatically only when **all** of these are true: * Your operating policy is **Active**, owner-approved, and **Approve Autonomous Sending** is on. * The objective is set to **Autonomous Send (Under Policy)** and is not paused. * The objective was approved under the **current** policy version. If you save changes to the policy, objectives must be approved again. * The objective has a target metric and an observation window of at least 3 days. * The campaign is not an experiment. * The channel is covered for automatic delivery, and holdout measurement is available when your holdout is above 0%. * The campaign stays inside the policy: permitted channel, permitted audience segment, and maximum discount. Even when all of the above is true, **email campaigns built from an approved email design are always reviewed by a person before they send.** See [Brand and email design](/guide/brand-and-email-design). If a campaign does not meet these conditions, it is held as a proposal for your review instead of sending. Revenue per recipient is the only metric whose results are measured end to end today. For objectives that target another metric, results cannot be measured automatically. ## Related * [Contact controls](/guide/contact-controls): set the limits agents must respect. * [Brand and email design](/guide/brand-and-email-design): how agents design emails within your brand. * [Dashboard Home and AI Copilot](/guide/dashboard-home-ai-copilot): approve pending proposals from Home. * [AI Agents](/guide/ai-agents): an overview of the agents in Tranthor. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.