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

# Chat agent: query your AR data with natural language

> Daylit's chat agent answers questions about invoices, customers, and aging in plain English — no SQL or filters needed — directly from your AR dashboard.

<Warning>
  **Coming soon:** The chat agent is still rolling out in Daylit. When you open chat today, you'll be connected to a real person on the Ops team — we'll get back to you in a few minutes. This page describes the planned self-serve experience.
</Warning>

The chat agent is a text-based conversational interface built directly into your AR dashboard. Instead of navigating menus or building report filters, you type a question and get an answer drawn from your live AR data. The agent understands plain English, so you don't need to know column names, filter syntax, or report structures — just ask what you want to know the way you'd ask a colleague.

## How to open a chat session

<Steps>
  <Step title="Click the chat icon">
    Find the chat icon in the top navigation bar or on the dashboard. Clicking it opens the chat panel.
  </Step>

  <Step title="Start a new session or continue an existing one">
    The panel shows your recent sessions in the sidebar. Click **New session** to start fresh, or select a previous session to pick up where you left off.
  </Step>

  <Step title="Type your question">
    Enter your question in the text field and press Enter. You'll see the agent's response stream in as it's generated — you don't have to wait for the full answer to appear.
  </Step>
</Steps>

## Suggested prompts

When you open a new session, the chat panel displays three suggested prompts to help you get started quickly. These defaults reflect the most common AR questions:

* "Which customers are overdue?"
* "Any disputes to watch?"
* "What follow-ups are pending?"

As you chat, the suggested prompts update based on the topic of your conversation. For example, if you ask about overdue balances, the agent may suggest "Show pending actions" or "Any disputes to watch?" as natural next questions.

## What data the chat agent can access

The agent has access to your full AR dataset for your company, including:

* **Customers** — open balances, overdue amounts, aging bucket breakdowns, risk signals, and communication history
* **Invoices** — invoice numbers, amounts, due dates, current status, and payment predictions
* **Aging data** — portfolio-wide and per-customer breakdowns across all aging buckets
* **AI insights** — conversation summaries, risk signals, payment date predictions, and recommended collection actions

If you @-mention a specific customer or invoice in your message, the agent fetches detailed records for that entity and includes them as additional context — so your answer is grounded in exact data, not just portfolio-level summaries.

## Example questions

**Portfolio overview**

* "Give me a summary of where my AR stands today."
* "How much is overdue in total, and how is it spread across aging buckets?"
* "What's my total AR balance?"

**Customer-specific**

* "What's the status of Mesa Valley Construction's account?"
* "Which customers have a 'ghosting' risk signal?"
* "Who has the largest overdue balance right now?"

**Invoice status**

* "What's the status of invoice 1042?"
* "Show me all invoices that are more than 60 days past due."
* "Which invoices have a dispute open?"

**Collection recommendations**

* "Who should I follow up with today?"
* "What action does the AI recommend for Apex Solutions?"
* "Which accounts have a promise to pay expiring this week?"

## Streaming responses

The chat agent's answers appear as they're generated — word by word — so you see the response building in real time rather than waiting for a complete answer. This streaming behavior also means you can start reading and processing the answer while the agent is still working on the rest.

If the agent needs to look up data (for example, fetching a customer's invoice list or pulling an AI insight), you'll briefly see a tool activity indicator before the response text begins.

## How sessions are saved and named

Each chat session is saved automatically. Once you've had at least one exchange with the agent, Daylit generates a session title based on what you discussed — for example, "Overdue balance review" or "Mesa Valley Construction follow-up." This title appears in the session sidebar so you can find previous conversations easily.

Sessions are private to your user account. Other team members on the same company cannot see your chat history.

## Starting a new session vs. continuing an existing one

* **New session** — the agent starts fresh with a full snapshot of your current AR data. Use this when you're switching topics or want a clean context.
* **Existing session** — the agent picks up the conversation history. Follow-up questions in an existing session don't require you to re-explain the context, since the agent already has the prior exchange.

<Tip>
  For unrelated questions, start a new session rather than continuing a long thread. Shorter, focused sessions make it easier to find specific conversations later.
</Tip>