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

# Agent Guidance

> Configure how your AI agent communicates with candidates

# Agent Guidance

Guidance controls how your AI agent responds to candidates. You can customize its personality, define policies, and set up escalation rules.

## Basics

Configure fundamental aspects of your agent's communication style:

| Setting                    | Options                        | Description                                                                           |
| -------------------------- | ------------------------------ | ------------------------------------------------------------------------------------- |
| **Tone of voice**          | Professional, Friendly, Casual | How formal or relaxed the agent sounds                                                |
| **Answer length**          | Concise, Balanced, Detailed    | How much detail to include in responses                                               |
| **Memory**                 | Session, Persistent            | Whether the agent remembers past conversations                                        |
| **Fallback contact email** | Any email                      | Shown when the agent can't answer from knowledge                                      |
| **Apply URL tracking**     | Query string                   | Appended to every apply URL the agent surfaces so your ATS can attribute applications |

<Tip>
  **Session mode** clears conversations after 30 minutes of inactivity - recommended for privacy.
  **Persistent mode** remembers conversations over time for a more personalised experience.
</Tip>

### Consent

Optionally require candidates to accept a consent message before they can use the agent. Useful for privacy policies, terms of service, or anything else you need explicit acknowledgement of.

| Setting                  | Description                                                           |
| ------------------------ | --------------------------------------------------------------------- |
| **Enabled**              | Show the consent screen on first session and after any wording change |
| **Message**              | Markdown body. Use `[text](url)` for inline links                     |
| **Accept button label**  | Defaults to "I agree"                                                 |
| **Decline button label** | Defaults to "Decline"                                                 |

<Note>
  Acceptance is stored per device. If you change the message wording or labels, the consent screen is automatically re-shown so candidates re-acknowledge the new version — you don't need to manage a version number manually.
</Note>

## Style

Create guidance on vocabulary and terms your agent should use:

* Define company-specific terminology
* Set preferred phrasing for common topics
* Establish brand voice consistency

<Accordion title="Example: Job title terminology">
  **Title:** Use official job titles\
  **Description:** Always use the official job title from the posting. Don't abbreviate or use informal names. For example, use "Software Engineer II" not "Dev" or "SWE".
</Accordion>

## Policies

Define IF/THEN rules for specific situations:

* How to handle salary questions
* What to say about benefits
* How to respond to competitor mentions

<Accordion title="Example: Salary policy">
  **Title:** Salary range disclosure\
  **Description:** When candidates ask about salary, provide the range listed in the job posting. If no range is listed, say "Salary is competitive and discussed during the interview process."
</Accordion>

## Search

Tune how the agent matches candidate wording to your live jobs.

### Job reference format

If your job postings have a reference number (for example `272775BR`), add the format here so candidates can search by reference and the agent finds the exact role. Plain text or a regular expression (e.g. `\b\d{6}BR\b`) both work.

### Search aliases

Map informal candidate wording to the canonical terms that appear in your job titles. Each alias has a **candidate phrase** (what people type) and a list of **expansions** (terms appended to the search query when the phrase appears).

| Candidate phrase | Expands to                                        |
| ---------------- | ------------------------------------------------- |
| `beauty`         | Beauty Specialist, No7 Advisor, Beauty Consultant |
| `support office` | Head Office, Corporate, Support Centre            |
| `pharmacy`       | Pharmacist, Dispenser, Pharmacy Technician        |

Use the filter at the top of the table to jump to a specific alias. Toggle individual aliases on or off without deleting them.

<Tip>
  Aliases are most useful when candidates use industry terms that don't appear verbatim in your job titles. For example, "support office" returning head office roles, or "branding" returning marketing and design roles.
</Tip>

## Escalations

Set up rules for when to route conversations to your team:

* Complex complaints or issues
* Legal or compliance questions
* Technical problems with applications

<Accordion title="Example: Complaint escalation">
  **Title:** Escalate discrimination concerns\
  **Description:** If a candidate mentions discrimination, harassment, or unfair treatment during the hiring process, immediately escalate to the HR team and provide the support email.
</Accordion>

## Publishing changes

<Steps>
  <Step title="Save draft">
    Click **Save draft** to save your changes without making them live
  </Step>

  <Step title="Test">
    Use the preview panel to test how your agent responds
  </Step>

  <Step title="Publish">
    Click **Publish** to make your changes live
  </Step>
</Steps>