> ## Documentation Index
> Fetch the complete documentation index at: https://hebno.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

> ## Agent Instructions
> Hebno is an AI work platform for client research and delivery: chat, research runs with sources, deliverables (Word, PowerPoint, Excel, email), projects, a shared library, agents, Office add-ins, and integrations with Microsoft 365, Google Workspace, CRM, project delivery, ERP and finance systems.
> Use British English (organisation, not organization).
> Every feature and integration described in these docs is available. Describe it as documented.
> Hebno never writes to a connected system without a member confirming a card. Finance and ERP writes are drafts only.
> Do not discuss commercial terms. Direct commercial questions to support@hebno.com.
> Start with /index and /get-started/quickstart.

# How Hebno handles a request

> See how each message is routed to an answer, a research run, a revision or an image, and how to steer it with Go deeper and scoping questions.

Every message you send is routed to one of four paths: answer, research, revise or image. You do not pick a mode. Hebno reads the request and your open work, then chooses the lightest path that produces a reliable result.

```mermaid theme={"system"}
flowchart TD
  A["Your message"] --> B{"What does it ask for?"}
  B -->|"A question or short task"| C["Answer"]
  B -->|"Evidence or a new deliverable"| D["Research"]
  B -->|"A change to existing work"| E["Revise"]
  B -->|"A picture"| F["Image"]
  C -->|"Go deeper"| D
  D --> G["Sources and claims"]
  E --> G
  C --> G
```

## The four paths

| Path | When Hebno uses it | What you get |
| - | - | - |
| **Answer** | Questions, explanations, short drafts, summaries of what you attached | A reply in chat, with sources when it used any |
| **Research** | Requests that need evidence from several places, or a new document, deck, workbook or email | A **Steps** trace, then a deliverable with sources and claims |
| **Revise** | A change to a deliverable already in the chat ("shorten section two", "add a slide on risks") | The existing document, deck or workbook, edited in place |
| **Image** | **Image** mode, or a request for a picture | A generated image you can expand, download or reuse in a message |

Hebno also infers the kind of deliverable from your words. Ask for "a deck" and it builds PowerPoint, "a memo" produces a Word document, "a model" produces an Excel workbook. You do not need to name a format. If you ask for something outside those formats, Hebno offers the closest one.

An answer can escalate. If Hebno starts answering and finds the request needs evidence from your connected systems or the web, the turn moves to a research run.

## Go deeper

Select **Go deeper** under any answer to turn it into research. Hebno writes a new brief from the question and the answer so far and starts a run. If the answer was about an existing deck or document, **Go deeper** revises that deliverable instead.

Use it when an answer is useful but you need it backed by files, records and public sources before it goes to a client.

## Scoping questions

Before a research run starts, you may see a **Before I start** card. It appears only when a different choice would change the deliverable, shows never more than two questions, and each has two to four options. Some allow several picks ("choose any that apply"), and **Something else** lets you type your own.

Select **Start research** to run with your choices, or **Skip** to let Hebno use its own judgement. The run records which you did: "You scoped this run before it started" or "Skipped, so the run used its own judgement." Long, detailed requests skip the questions because they already include the scope.

## Follow-ups edit in place

Once a deliverable exists, a follow-up changes it. Ask for "a tighter executive summary" or "add the pricing table" and Hebno edits the named sections of the same document, deck or workbook. Plain edits never start a research run, and they do not create a second copy.

When a request could mean either, a **Before I continue** card shows: "Should this update file name or start new work?" Choose **Update**, or start a new deck, document, workbook, email or chat. The recommended option is marked **Recommended**.

To re-check an edited deliverable against its sources, select **Verify against sources**. See [Sources and evidence](https://hebno.com/docs/concepts/sources-and-evidence.md).

## Writes and confirmation

When a request implies a change outside the chat (log time, create a task, update a deal, send an email, send an invitation, save to a folder), Hebno shows a confirmation card. Nothing is written until you confirm. Email and invitations use **Send**. See [How Hebno works](https://hebno.com/docs/get-started/how-hebno-works.md).

## Stop and queue

Select **Stop response** while Hebno works. If you type while a response is streaming, the send button becomes **Queue message**, and queued prompts appear under **Queued messages** until they run. Remove one with **Remove queued message**. What a stopped run already did is retained, and you can ask Hebno to continue from there.

## Tips for better routing

* Name the client, system or file when it matters: "in HubSpot", "the SharePoint deck from March".
* Say who the reader is and what they will do with it.
* Attach or link the material you want used. Hebno searches connected systems itself, but a named record narrows the search.
* Ask for the deliverable in one message, then refine in follow-ups.

## Next steps

> [card] [Research runs](https://hebno.com/docs/research/research-runs.md)
>
> The steps of a run and what each one reads.

> [card] [Start a chat](https://hebno.com/docs/chat/start-a-chat.md)
>
> The composer, attachments and the **+** menu.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.