<!-- Generated from the canonical Teloring documentation source. Do not edit this file. -->

Canonical page: https://docs.teloring.com/docs/product/ai-world
Last modified: 2026-08-20T20:49:57.000Z

# AI World

**AI World** is the control room for every artificial‑intelligence feature in Teloring. It's a single page of on/off switches — one per capability — that decide how much AI helps your team across conversations, the knowledge base, and analytics.

Turn a switch **on** and that capability becomes available everywhere it belongs: the reply composer, the AI Copilot panel, the resolve action, the Knowledge Base page, or the Analytics report builder. Turn it **off** and it disappears again. Nothing here changes what your agents *can* do by hand — it only decides where AI lends a hand.

Open **AI World** from the **Tools** section of the sidebar.

![The AI World page — the grid of AI feature switches](pathname:///img/screenshots/product/ai-world/ai-world-overview.png)

## Key facts

| Fact | Meaning |
| --- | --- |
| Account‑wide | Every switch applies to your whole business. When an admin turns a feature on, it's on for **all** agents in the account. |
| Admins change it, everyone benefits | Only account **administrators** can flip a switch. Agents can view the page but the toggles are read‑only for them. |
| Off by default | Every feature starts **off**. Nothing sends data to AI until an admin turns it on. |
| Saves instantly | There is no Save button. Each switch saves the moment you flip it and shows a short **Saving… → Saved!** status on the card. |
| One master switch | **AI Copilot** is the master switch for the six live‑conversation helpers. Those six do nothing unless AI Copilot is also on — see [The AI Copilot dependency](#the-ai-copilot-dependency). |
| Your data stays yours | AI only ever reads the conversations, files, and records of your own account. Features are fully account‑isolated. |
| Some cards are on the roadmap | A few switches are shown for what's coming and are **not active yet** — see [Coming soon](#coming-soon). |

## Who uses AI World

| Role | What they do here |
| --- | --- |
| Admins | Decide which AI features the account uses, and turn them on or off. |
| Agents & managers | Benefit from whatever is enabled — Copilot tips, rephrase, summaries, and more — but cannot change the switches. A lock notice appears at the top of the page for them. |

:::note
If you open AI World and the switches won't move, you're signed in as a non‑admin. Ask an account administrator to change the settings for you.
:::

## How the page works

Every capability is shown as a **card** with an icon, a name, a one‑line description, and a switch on the right.

![A single feature card, switched on](pathname:///img/screenshots/product/ai-world/feature-card-active.png)

| Element | What it means |
| --- | --- |
| Card icon & title | The feature's name and a plain‑language description of what it does. |
| Switch (right side) | Click to turn the feature on or off. |
| **Active** highlight | When a feature is on, its card gets a teal border and subtle tint, so you can see at a glance what's enabled. |
| Status text | A small label next to the title flashes **Saving…**, then **Saved!** (or **Failed to save** if something went wrong). |

To enable a feature: click its switch **on**. To disable it: click it **off**. That's the whole workflow — the change takes effect across the account within moments.

## The AI Copilot dependency

Six features are **helpers that run inside AI Copilot**. They only work when the **AI Copilot** switch is also on:

- Help & Tips
- Smart Subject Detection
- Auto‑Assign Labels
- Detect Churn Risk
- Detect Upsale Opportunity
- Auto Mark Urgent

Think of **AI Copilot** as the engine and these six as options you bolt onto it. If AI Copilot is **off**, turning any of the six on has no effect. If AI Copilot is **on**, each of the six you enable adds one more thing Copilot watches for while a conversation is live.

:::tip
A good starting point for most teams: turn on **AI Copilot**, then enable **Help & Tips**, **Smart Subject Detection**, and **Auto‑Assign Labels**. Add churn, upsale, and urgency detection once your team is comfortable.
:::

---

## Feature reference

The sections below cover every **live** feature — what it does, where you'll see it, and what it needs to work.

### Live‑conversation assistance (AI Copilot)

These features act while agents are chatting with customers. Copilot analyzes each new incoming customer message in the background and surfaces its findings in the **AI Copilot** panel on the right of the conversation. See [AI Copilot in conversations](/docs/getting-started/conversations#ai-copilot) for the agent‑side view.

Copilot only runs on conversations that are **open or pending** and **not** already handled by a Studio flow or an AI agent.

#### 🧠 AI Copilot

The master switch for live assistance. When on, Copilot reads each incoming customer message and can post insights into the Copilot panel. On its own it does nothing visible — pair it with the helpers below and with **Knowledge Base** (for answers).

| | |
| --- | --- |
| **Where you see it** | The AI Copilot panel in the right sidebar of a conversation. |
| **Requires** | Nothing, but only useful with at least one helper below turned on. |

#### 💡 Help & Tips

Copilot offers real‑time suggestions and coaching — how to respond, what the customer seems to need, and what tone fits. Tips can appear several times as a conversation develops.

| | |
| --- | --- |
| **Where you see it** | As Copilot tips in the AI Copilot panel. |
| **Requires** | **AI Copilot** on. |

#### 🎯 Smart Subject Detection

Copilot writes a short, clear subject line for the conversation once it understands the topic.

| | |
| --- | --- |
| **Where you see it** | The conversation's subject updates automatically; Copilot notes that it set it. |
| **Requires** | **AI Copilot** on. |
| **Good to know** | Only sets the subject when the conversation **doesn't already have one**. It never overwrites a subject you set yourself. |

#### 🏷️ Auto‑Assign Labels

Copilot tags the conversation with relevant category labels (for example `billing, refund` or `technical, bug`), creating them in your account if they don't exist yet.

| | |
| --- | --- |
| **Where you see it** | Labels appear on the conversation; Copilot notes what it added. |
| **Requires** | **AI Copilot** on. |
| **Good to know** | Only labels a conversation that has **no labels yet**, up to 5 labels. It won't touch conversations you've already labelled. |

#### 🚪 Detect Churn Risk

Copilot watches for signs that a customer wants to cancel, leave, or is deeply dissatisfied, and raises a clear alert describing what triggered it.

| | |
| --- | --- |
| **Where you see it** | A churn alert in the AI Copilot panel. It can also start a **Studio** automation (a churn‑risk signal). |
| **Requires** | **AI Copilot** on. |

#### 💰 Detect Upsale Opportunity

Copilot spots when a customer shows interest in buying, upgrading, or spending more, and flags the opportunity for the agent.

| | |
| --- | --- |
| **Where you see it** | An upsale alert in the AI Copilot panel. It can also start a **Studio** automation (an upsale signal). |
| **Requires** | **AI Copilot** on. |

#### 🔥 Auto Mark Urgent

Copilot assesses how urgent a conversation is and sets its priority — `urgent`, `high`, `medium`, or `low`.

| | |
| --- | --- |
| **Where you see it** | The conversation's priority changes automatically. |
| **Requires** | **AI Copilot** on. |
| **Good to know** | Only adjusts priority while it's still at the default **medium**. It won't override a priority you set by hand. |

### Writing help

#### ✍️ Rephrase Messages

Gives agents a one‑click **✨ rephrase** button in the reply composer. Before sending, an agent can rewrite their draft in several ways.

![The rephrase picker in the reply composer](pathname:///img/screenshots/product/ai-world/rephrase-picker.png)

| Rephrase option | What it does |
| --- | --- |
| Improve | Makes the text clearer, more polished, and professional. |
| Fix spelling & grammar | Corrects mistakes without changing meaning or tone. |
| Expand | Adds detail and elaboration. |
| Shorten | Trims to the essentials. |
| Friendly | Rewrites in a warm, approachable tone. |
| Formal | Rewrites in a professional, business tone. |
| Simplify | Uses short sentences and simple words. |
| Translate | Translates the reply into the **customer's** language, detected from their last message. |

| | |
| --- | --- |
| **Where you see it** | The ✨ button next to the reply composer's formatting tools. |
| **Requires** | Nothing else — works on its own. |
| **Good to know** | The AI keeps the agent's original language for every option except **Translate**. Agents always review the result before sending. |

### When a conversation is resolved

These two features run automatically the moment an agent marks a conversation **Resolved**. They read the conversation (skipping private notes) and act in the background.

#### 📝 Conversation Summary

Generates a concise 2–4 sentence summary of the conversation — the topic, what the customer needed, and how it was resolved — written in the **same language** as the conversation.

| | |
| --- | --- |
| **Where you see it** | Added as a **private note** (🤖 AI Summary) at the end of the conversation, visible to your team but never to the customer. |
| **Requires** | Nothing else. |

#### 💭 Feelings & Emotions

Reads the customer's overall mood across the conversation and records it on a 20‑point scale — from *satisfied*, *happy*, and *appreciative*, through *neutral*, to *frustrated*, *angry*, and *threatening*.

| | |
| --- | --- |
| **Where you see it** | Stored on the conversation (visible in resolved‑conversation views and available to **Analytics** and **Studio**). |
| **Requires** | Nothing else. |
| **Good to know** | A customer‑emotion change can start a **Studio** automation — useful for escalating an angry ending, for example. |

### Knowledge & media

#### 📚 Knowledge Base

Unlocks the **Knowledge Base** page, where you upload, scan, and store documents for AI to reference. It also lets **AI Copilot** answer agents' questions from your approved knowledge.

| | |
| --- | --- |
| **Where you see it** | The Knowledge Base page (in the sidebar) becomes usable; Copilot can search it. |
| **Requires** | Nothing to enable it. Copilot's knowledge answers need **AI Copilot** on as well. |
| **Good to know** | While this is off, the Knowledge Base page shows a "feature disabled" message instead of your files. |

#### 🖼️ Image Detection

For every image a customer sends, AI looks at it and describes what's in it — objects, brands, documents, and any visible text.

| | |
| --- | --- |
| **Where you see it** | The description is attached to the incoming image in the conversation, and appears as soon as the analysis is ready. |
| **Requires** | Nothing else. |

### Automation

#### 🎨 AI Studio

Adds a second way to create a Studio flow: **Hermes**, an AI flow builder that interviews you in plain language and then lays out the whole automation — blocks, settings, and connections — on your canvas.

![The New Flow dialog with the two build-mode cards](pathname:///img/screenshots/product/studio/ai/new-flow-choose-mode.png)

| | |
| --- | --- |
| **Where you see it** | Studio → **New Flow** → *How would you like to build it?* → **Use AI to build**. |
| **Requires** | Nothing else. It helps to connect the inboxes, teams, and CRM objects you want to automate first, because Hermes offers your real ones by name. |
| **Who can use it** | Any agent with **Studio → Create**. Flipping this switch is a different permission — **AI World → Update**. See [Roles and Permissions](./roles/overview.md). |
| **Good to know** | Hermes always produces a **draft**. Nothing goes live until someone publishes it from the editor. Building the flow is a one-way step that ends the conversation; the flow is then edited by hand like any other. |

While this switch is off, the **Use AI to build** card still appears in Studio but is greyed out, with a link back to this page.

📖 Full guide: [Build a flow with AI (Hermes)](/docs/product/studio/ai-flow-builder).

### Analytics

#### 📊 AI Reports & Analytics

Adds a **✨ Use AI to build the report** button to the Analytics report builder, so anyone can describe a report in plain language and let AI fill in the fields.

| | |
| --- | --- |
| **Where you see it** | The **✨ Use AI to build the report** button in [Analytics](/docs/product/analytics#build-a-report-with-ai). |
| **Requires** | Nothing else. When off, the button is hidden. |

---

## Coming soon

A few switches appear on the AI World page to show what's on the roadmap. They are **not active yet** — turning them on does not change anything in the product today. We're listing them here so you know what each one is intended to do.

:::info
These features are in development. Leave them off until this guide marks them as live.
:::

| Feature | Planned behaviour |
| --- | --- |
| 🌍 **Conversation Translation** | Translate whole conversations in real time to communicate across languages. |
| 🎙️ **Conversation Transcription** | Turn voice messages and audio into text automatically. |
| 🧩 **Enrich Knowledge Base** | Grow the knowledge base automatically from the solutions your agents give in real conversations. |
| 🛡️ **Prevent Bad Language** | Stop agents from sending inappropriate language to customers. |
| 🔗 **Detect Similar Issues** | Spot when several recent conversations report the same problem and alert your team. |
| 🔐 **AI Security** | Detect security risks and signs of compromise in real time. |
| 💡 **AI Suggestions** | Surface AI‑only insights and suggestions on the main dashboard. |
| 🛡️ **AI File Protection** | Scan every incoming and outgoing file and block malicious ones. |
| 🎓 **Auto Agent Skills** | Detect each agent's strengths automatically from their conversations and performance. |

## Frequently asked questions

**Do I have to turn everything on?**
No. Start with the features you need. Every switch is independent (except the six that need AI Copilot — see [above](#the-ai-copilot-dependency)).

**Will customers know AI is involved?**
Customer‑facing help is always in the agent's hands. Rephrase suggestions, Copilot tips, summaries, and emotion scores are shown to your **team** only — nothing is sent to the customer automatically. Summaries are saved as private notes.

**Can agents change these settings?**
No — only account administrators. Agents see the page in read‑only mode with a lock notice.

**A switch flipped back off after I clicked it.**
Enabled features stay on. If a switch won't stay on, it may be one of the roadmap items in [Coming soon](#coming-soon), which aren't active yet — check with your Teloring contact.

**Where do I actually see the AI at work?**
In the [conversation workspace](/docs/getting-started/conversations#ai-copilot) (Copilot panel, rephrase, image descriptions), on resolved conversations (summaries and emotions), on the Knowledge Base page, in [Studio](/docs/product/studio/ai-flow-builder) when you create a flow, and in [Analytics](/docs/product/analytics#build-a-report-with-ai).
