> ## Documentation Index > Fetch the complete documentation index at: https://wiredesk.ai/docs/llms.txt > Use this file to discover all available pages before exploring further. # Help while you reply > A summary when you take over, replies drafted from your agent's knowledge, saved replies, and translation, all in the reply box on a chat, email or WhatsApp conversation. When you take over a chat, email or WhatsApp conversation, the reply box under **Reply as owner** has four things to make the reply faster. None of them sends anything. They write into the box, you read and change the text, and the customer gets nothing until you select **Send**. Everyone in the workspace can use them, whatever their [role](https://wiredesk.ai/docs/account/team-and-roles.md). ## The summary At the top of the reply box, **So far** sums up the conversation for whoever is taking it over: | Line | What it holds | | - | - | | **They want** | One sentence on what the customer is after. | | **The agent already** | What the agent answered, offered or looked up in its knowledge, and what came of it. | | **Details collected** | The name, email and phone number the agent saved (from the contact, a lead or a message it took), plus things the customer gave you need to act on, such as an order number. | It is written when the customer asks for a person, so it is usually there before you open the conversation. If the conversation was taken over without anyone asking, it is written when you open the page. On any other chat, email or WhatsApp conversation, select **Summarise the conversation so far**. When the customer writes again after the summary was written, its button reads **New messages since — refresh**. Select it to rewrite the summary from the conversation as it stands. ## Draft reply **Draft reply** writes a reply to the customer's latest message from: * the agent's knowledge, searched with the customer's last two messages * the agent's instructions, or the channel's own if that channel has stopped inheriting them * the whole conversation so far Under the draft, **From** lists the knowledge sources it used. Hover over one to see the passage. If nothing in the knowledge matched, it says so: check every fact before you send. Where the knowledge does not cover what the customer asked, the draft puts what you need to fill in inside square brackets, for example *\[check the refund date on the order]*, instead of making something up. Replace the brackets before you send. A draft replaces what was in the box. **Undo draft** puts back what you had typed. ## Saved replies A saved reply is an answer your team sends often, written once. Select **Saved replies** in the reply box and pick one. It is added below anything you have already typed, with its variables filled in. Once there are five or more, you can search them by title or text. Saved replies belong to the workspace and everyone in it can use them. The owner and admins add, change and remove them under **Settings** › **Saved replies & translation**, or from **Manage saved replies** at the foot of the picker. ### Variables | Variable | Fills in with | | - | - | | `{{contact.name}}` | The customer's name, if the agent learned it | | `{{contact.first_name}}` | The first word of it | | `{{contact.email}}` | Their email address | | `{{contact.phone}}` | Their phone number | | `{{me.name}}` | Your name | | `{{me.first_name}}` | Your first name | | `{{agent.name}}` | The AI agent's name | Add a fallback after a `|` for when the value is not known: ``` Hi {{contact.first_name | there}}, ``` reads "Hi Jane," when the name is known and "Hi there," when it is not. Without a fallback an unknown value is left blank. A variable that is not in the list, such as a typo, is left exactly as written so you can see it before you send. A saved reply can be up to 2,000 characters, the same as a reply, and a workspace can keep 200. If adding one to what you have typed would take the reply past 2,000 characters, it is not added and the box says why: shorten what is there first. As a reply nears the limit, the box counts its characters, and **Send** stays off while it is over. ## Translation Under **Settings** › **Saved replies & translation**, **Your team works in** sets the language your team reads and writes. It is English to start with, and the owner and admins can change it. Summaries and drafted replies are written in it. Wiredesk learns which language a customer writes in when it writes the summary, drafts a reply, or translates. * **The conversation.** Above the transcript, **Translate to** *your language* shows the conversation in your team's language: the customer's messages and the replies to them, which the agent wrote in the customer's language. Each translated line is marked **Translated from** *their language*; a line already in your language is left as it is. New messages are translated as they arrive. A long thread is translated a few messages at a time; if some of them could not be translated, the bar says how many and **Try the rest again** retries them. **Show original** switches back. The transcript itself is never changed. * **Your reply.** Once the customer's language is known and is not your team's, the reply box offers **Translate to** *their language*. Before that it offers **Translate to their language**, which works the language out from their messages. The translation replaces your text in the box, so you can check it before you send. **Undo translation** puts back what you wrote. Translations often run longer than the original; if one would be over 2,000 characters, it is not put in the box and you are asked to shorten your reply and translate again. The agent itself already answers customers in their own language. Translation is for the replies you write yourself. ## Related * [Conversations](https://wiredesk.ai/docs/workspace/conversations.md) * [Handover](https://wiredesk.ai/docs/concepts/handover.md) * [Team and roles](https://wiredesk.ai/docs/account/team-and-roles.md) This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.