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

> ## Agent Instructions
> This is the Journeyfront Help Center for hiring teams. Journeyfront is an applicant tracking system (ATS). Use the Help Center's terms: job req, Candidate Profile (the person), Candidate Application (one person on one job req), hiring class. Link to the article you used in your answer.

# AI agents

> Set up AI agents that follow up with stalled candidates, with your team approving what they send.

> [!NOTE]
>
> **Who can do this:** Four permissions, set on each security role:
>
> * *Manage AI Agents*: create and set up agents.
> * *Assign AI Agents*: turn agents on for a job req and assign candidates to them. Turning an agent on for a job req also needs *Manage Job Req Settings*.
> * *Approve AI Agent Actions*: approve, edit or reject the messages agents propose.
> * *View AI Agent Tasks*: see what agents have done, read-only.
>
> Assigning, approving and viewing only cover the job reqs you have access to.
>
> **Availability:** AI agents need to be turned on for your account. Ask your Customer Success Manager. Agents use [AI credits](#ai-credits).

Candidates stall all the time: an assessment left half done, an interview invite never booked. An **AI agent** is a named teammate that follows up for you. It looks at each candidate you assign to it, works out whether they have something left to do, and sends a short, personal email, text or WhatsApp message to get them moving again. Most of the time, the right call is to do nothing, and that's what it does.

People stay in control. You choose which candidates an agent works on and which channels it can use, and for each channel, whether it sends on its own or waits for someone on your team to approve every message. An agent can only send messages: it can't move, reject or score anyone. Every action it takes is logged with its reasoning.

## How an agent works

On its schedule, and whenever something happens to a candidate (they apply, finish an activity or move to a new step), the agent looks at each candidate assigned to it:

* **Where they are:** their current step, and which activities in it are theirs to do, like screening questions, an assessment, a typing test, an interview they've been invited to book, or an onboarding form that's been sent to them.
* **What's already been said:** every email and text in their history, from anyone on your team, your automation rules or the agent itself.

Then it decides to send a nudge, hold a message for your approval, hand the candidate off to your team, or do nothing. It's built to follow up promptly when a candidate first stalls, space later follow-ups further apart, and stop after several tries with no response.

## What an agent never does

These rules are always on and can't be turned off. Most of them are also listed under **Guardrails** in every agent's settings.

* It only acts on candidates you've assigned to it.
* It never messages a closed application: hired, rejected or withdrawn. It never messages a blocked candidate.
* It only texts or WhatsApps candidates who have opted in to that channel. Email doesn't need an opt-in.
* It's outbound only. Sending messages is the only action it takes. It can't change a candidate's step, reject them, score them or assign them.
* It works only from hiring workflow data, never protected characteristics like age, gender or race.
* It leaves offers and background checks alone. To remind a candidate about an offer, see [E-signatures](https://help.journeyfront.com/offers/e-signatures.md).
* Every action is logged, with the agent's reasoning.

## Before you start

1. **Turn on AI agents.** Ask your Customer Success Manager.
2. **Give people the permissions.** Go to **Settings**, then **Security**, and turn on the AI Agents permissions for the right roles. See [Security roles](https://help.journeyfront.com/account-setup/security-roles.md).
3. **Set up texting, if you want it.** Email works right away. To let an agent send texts or WhatsApp messages, your account needs a number for that channel in **Settings**, then **Phone Numbers**, and WhatsApp messages also need approved WhatsApp templates. See [Text messages](https://help.journeyfront.com/communication/text-messages.md) and [WhatsApp](https://help.journeyfront.com/communication/whatsapp.md).
4. **Check your AI credits.** See [AI credits](#ai-credits).

## Create an agent

**1. Open AI Agents**

Go to **Settings**, then **AI Agents**, and click **New agent**.

![Settings, AI Agents with the agents Riley (Active) and Morgan (Paused) in the list, and the New agent button highlighted next to Agent activity](https://mintcdn.com/journeyfront/PQrdl_4IlrPf8aOR/images/workflows/ai-agents/ai-agents-new-agent.png?fit=max\&auto=format\&n=PQrdl_4IlrPf8aOR\&q=85\&s=e704bc298c2ab84aaf2320a7188e265b)

**2. Choose the type and name it**

Choose the **Type**: **Candidate Management** sends candidates reminders and nudges to finish the steps that are theirs to complete. Give the agent a **Name** and, if you like, a short **Description** for the agents list.

**3. Click Create agent**

The agent is created **Paused**, and its settings open so you can finish setting it up.

![The New agent panel in Settings, AI Agents, with Type set to Candidate Management and its description, the Name Riley, a short Description, and the Create agent and Cancel buttons](https://mintcdn.com/journeyfront/Ki8OScNKQAoeTHdj/images/workflows/ai-agents/gen-new-agent.png?fit=max\&auto=format\&n=Ki8OScNKQAoeTHdj\&q=85\&s=b5b39b46ef146f2759cd38ad8ad733fb)

> [!TIP]
>
> Name your agent like a teammate, for example "Riley" or "US Sales Candidate Manager". The name appears on every message it sends and every entry in a candidate's history, so your team always knows what the agent did.

## Set up the agent

Click an agent in the list to open its settings.

![Settings, AI Agents with the agent Riley highlighted in the list, showing its type Candidate Management, its description and the Active status](https://mintcdn.com/journeyfront/PQrdl_4IlrPf8aOR/images/workflows/ai-agents/ai-agents-open-agent.png?fit=max\&auto=format\&n=PQrdl_4IlrPf8aOR\&q=85\&s=3230ceb6c45350c9ac0394955f8cb387)

* **Name** and **Description.**
* **Instructions:** your own guidance on top of the agent's built-in role, such as tone, things to mention, or things to avoid. For example: "Keep messages warm and concise. Mention our 24-hour response promise." Instructions can narrow what the agent does; they can't switch off its guardrails.
* **Tools:** what the agent is allowed to do. **Read candidate process** is always on. Turn **Send SMS**, **Send WhatsApp** and **Send email** on or off, and choose when each needs your approval (see [Approval settings](#approval-settings)).
* **Send settings for each channel:**
  * **Send SMS** and **Send WhatsApp:** **Send from**, the number the agent messages from.
  * **Send email:** **From name**, the sender name candidates see (your company name if you leave it empty), and **Reply-to email**, where candidate replies go: the default address, or a **Custom email address**.
* **How often should the agent run:** **Every hour**, **Every 2 hours**, **Every 4 hours** (the default), **Every 6 hours**, **Every 12 hours** or **Once a day**. This is how often it looks over its candidates, not how often it messages them.

Click **Save changes**. When you're ready for the agent to start working, turn on the switch at the top of its settings. It changes from **Paused** to **Active**.

> [!TIP]
>
> Set **Reply-to email** to an inbox your team checks. The default is Journeyfront's support address, so without a custom address, candidates' replies to the agent's emails won't reach you.

### Approval settings

For each channel, **When should this need your approval?**

![Settings, AI Agents with the settings for an agent named Riley open: Name, Instructions, the Tools list with Send email turned on and Always ask chosen under When should this need your approval, the From name and Reply-to email settings, How often should the agent run set to Every 4 hours, and the Guardrails list](https://mintcdn.com/journeyfront/Ki8OScNKQAoeTHdj/images/workflows/ai-agents/gen-agent-settings.png?fit=max\&auto=format\&n=Ki8OScNKQAoeTHdj\&q=85\&s=eae396a0df2c9c85368e742bfd806984)

| Setting | What happens |
| - | - |
| **Don't ask** | The agent sends on its own, without checking in first. Every message still appears in its activity log. |
| **Ask when unsure** | The agent sends on its own when it's confident, and holds the message for you when it isn't sure. This is the default. |
| **Always ask** | The agent drafts each message and holds it for your approval. Nothing sends until someone approves it. |

Until you click **Save changes** on a new agent's settings, it holds every message for approval, the same as **Always ask**.

> [!TIP]
>
> Start with **Always ask** on every channel and a small group of candidates. Once your team trusts the drafts, move email to **Ask when unsure** and keep texts on **Always ask**, or loosen further as you go. You can change the setting at any time.

## Turn an agent on for a job req

An agent only works on job reqs where it's turned on.

**1. Open the job req's AI Agents settings**

Open the job req and go to **Setup**, then **AI Agents**. Each agent shows **Enabled** or **Off** for this job req.

![A job req's Setup tab with AI Agents selected and the agent list highlighted: Riley is Enabled and Morgan is Off for this job req](https://mintcdn.com/journeyfront/PQrdl_4IlrPf8aOR/images/workflows/ai-agents/ai-agents-job-req-setup.png?fit=max\&auto=format\&n=PQrdl_4IlrPf8aOR\&q=85\&s=8178ccd0375b5daf07c9df500fd927ff)

**2. Turn the agent on**

Click the agent and turn on **Enabled on this requisition**.

**3. Choose auto-assign (optional)**

Turn on **Auto-assign new candidates** to assign every new applicant to this job req to the agent automatically.

![A job req's Setup tab with AI Agents selected: Riley is marked Enabled and Morgan Off, and Riley's panel shows the Enabled on this requisition and Auto-assign new candidates switches turned on](https://mintcdn.com/journeyfront/Ki8OScNKQAoeTHdj/images/workflows/ai-agents/gen-job-req-agent.png?fit=max\&auto=format\&n=Ki8OScNKQAoeTHdj\&q=85\&s=859759cb72316c4f9ff0a8058deff77e)

## Assign candidates

An agent only works on the candidates assigned to it. A candidate can have more than one agent.

* **One candidate, from the list:** on the job req's **Candidates** tab, click the sparkle in the candidate's **AI Agent** column and pick an agent. Pick it again to unassign it. If you don't see the column, add it with **Edit Columns**.
* **One candidate, from their application:** open the Candidate Application and click the **AI Tasks** tab. **AI agents on this candidate** lets you add or remove agents.
* **Many candidates at once:** on the job req's **Candidates** tab, switch to the classic layout with **Back to Classic**, select the candidates, click **Assign AI Agent**, choose the agent, and click the **Assign to** button, which shows how many candidates you selected.
* **Every new candidate:** turn on **Auto-assign new candidates** for the job req.

![A job req's candidate list with the AI Agent column: one candidate shows Riley's avatar, and the sparkle on another candidate's row is open to the AI Agents menu listing Riley and Morgan](https://mintcdn.com/journeyfront/Ki8OScNKQAoeTHdj/images/workflows/ai-agents/gen-assign-agent.png?fit=max\&auto=format\&n=Ki8OScNKQAoeTHdj\&q=85\&s=6e06d583fcd799b10f2ccb434c28f502)

Only agents turned on for the job req can be assigned. An agent that's still assigned to a candidate but no longer available, for example because it was turned off for the job req, is marked **Off**, so you can still unassign it.

## Review and approve

Everything agents do, or want to do, is on the **AI Agent Tasks** page. Click the sparkle icon in the top navigation (a number shows how many messages are waiting), or **Agent activity** in **Settings**, then **AI Agents**.

![The AI Agent Tasks page with the sparkle icon highlighted in the top navigation, its badge showing 2 messages waiting, and two emails from Riley on the Awaiting approval tab](https://mintcdn.com/journeyfront/PQrdl_4IlrPf8aOR/images/workflows/ai-agents/ai-agents-tasks-nav.png?fit=max\&auto=format\&n=PQrdl_4IlrPf8aOR\&q=85\&s=f566c031c4f7e58debd7cb6b927ee867)

**1. Pick an agent**

The list on the left shows **All agents** and each agent, with how many of its messages are waiting. Narrow the list with **Search candidate**, **Job Req** and **Task type** (Email, Text or WhatsApp).

**2. Open a message**

On the **Awaiting approval** tab, click a task to see **Why the agent recommends this** and the draft.

**3. Edit it if you like**

Change the **Message**, and for emails the **Subject**, **From name** and **Reply-to email**. Add an **Approval note (optional)** for the record. **Reset to draft** puts back the agent's version.

**4. Approve or reject**

Click **Execute** to send the message, or **Reject** to drop it. To decide on several at once, tick them and click **Approve selected** or **Reject selected**.

![The AI Agent Tasks page with the agent Riley selected and three emails on the Awaiting approval tab; one is expanded to show Why the agent recommends this and the editable From name, Reply-to email, Subject, Message and Approval note, with Reject and Execute buttons on the row](https://mintcdn.com/journeyfront/Ki8OScNKQAoeTHdj/images/workflows/ai-agents/gen-agent-tasks.png?fit=max\&auto=format\&n=Ki8OScNKQAoeTHdj\&q=85\&s=ed67e9133af58a15ba0a0c14078e008a)

The other tabs show what happened next:

| Tab | What's in it |
| - | - |
| **Awaiting approval** | Messages waiting for someone to approve or reject them |
| **Completed** | Messages that were sent, whether the agent sent them on its own or someone approved them, and candidates the agent handed off to your team, marked **Escalated** with its reason |
| **Rejected** | Messages someone rejected, with who rejected them and when |
| **Failed** | Messages that were approved but couldn't be sent, with the reason |

While a message for a candidate is waiting for approval, the agent doesn't propose another one for them. The same list for one candidate is on the **AI Tasks** tab of their Candidate Application, with a count of messages waiting.

People with *View AI Agent Tasks* but not *Approve AI Agent Actions* see the same pages without the approve and reject controls.

> [!TIP]
>
> Make reviewing the queue part of the daily routine, like checking email. A message that sits waiting for a day loses most of its value, and the agent won't follow up with that candidate until it's decided.

## What candidates see

Candidates get an ordinary email, text or WhatsApp message from your company: the email from the agent's **From name**, texts from your number. Messages usually include a link straight back to what they need to do, like their application or their interview booking link.

Every message an agent sends appears in the candidate's history on the **Comments & Activities** tab, with the agent's name and avatar, just like messages from your team. Messages waiting for approval don't appear until they're sent.

Agents don't hold conversations. Candidate replies come to your team as usual: texts in the message panel, and emails at the **Reply-to email** address. When a candidate needs a person, for example because they replied with a question, the agent is built to hand them off to your team instead of answering.

## AI credits

Each message an agent lines up uses one AI credit, whether it sends on its own or waits for approval. Rejecting a message doesn't give the credit back. When an agent looks at a candidate and decides nothing is needed, or hands a candidate off to your team, no credit is used.

When your AI credits run out, agents stop lining up new messages until more credits are added. To see your balance and which agent is using credits, go to **Settings**, then **Add-Ons**, then **AI Credits**. See [Add-ons and credits](https://help.journeyfront.com/account-setup/add-ons.md).

## AI agents and automation rules

They work well together.

| | Automation rules | AI agents |
| - | - | - |
| **Decides** | A fixed rule: when this happens, do that | Looks at each candidate and decides what helps |
| **Message** | The same template for everyone | Written for each candidate's situation |
| **Timing** | Every time the rule matches | Chooses when, and often chooses not to send |
| **Actions** | Can move, reject, assign and message candidates | Can only send messages |
| **Oversight** | Runs on its own | Reasoning on every action, with optional approval |

Use automation rules for the steps every candidate should go through the same way, and agents for the follow-ups that need judgment. See [Automation rules](https://help.journeyfront.com/workflows/automation-rules.md).

## FAQ

> [accordion] Can an agent reject or move candidates?
>
> No. Sending messages is the only thing it can do. When a candidate needs a decision, it's built to hand them off to your team, and you'll see them on the **Completed** tab marked **Escalated**.

> [accordion] Will an agent message candidates we've hired or rejected?
>
> No. It never messages a hired, rejected or withdrawn application.

> [accordion] Will it text candidates who haven't opted in?
>
> No. Texts and WhatsApp messages only go to candidates who opted in to that channel. Email doesn't need an opt-in.

> [accordion] Why isn't my agent doing anything?
>
> Check that:
>
> * The agent is **Active**, not **Paused**.
> * It's **Enabled on this requisition** for the job req.
> * Candidates are assigned to it.
> * At least one send tool is on, and for texts, the candidates have opted in.
> * The candidates have something left to do in their current step. Waiting on your team isn't something to nudge about.
> * There isn't a message for them already waiting on the **Awaiting approval** tab.
> * You have AI credits left.
>
> Remember that doing nothing is often the right call. Every message it does send is on the **AI Agent Tasks** page.

> [accordion] Can I change a message before it goes out?
>
> Yes, when it's waiting for approval. Edit it on the **AI Agent Tasks** page or the candidate's **AI Tasks** tab, then click **Execute**.

> [accordion] How often will a candidate hear from the agent?
>
> There's no fixed schedule. The agent reads each candidate's history, including messages from your team, and is built to space follow-ups further apart over time and to stop when a candidate doesn't respond. To control every message, use **Always ask**.

> [accordion] Why don't I see AI Agents?
>
> AI agents need to be turned on for your account, and each part needs its permission: **Settings**, then **AI Agents** for *Manage AI Agents*, the job req's **Setup**, then **AI Agents** and the **AI Agent** column for *Assign AI Agents*, and the **AI Agent Tasks** page for *View AI Agent Tasks* or *Approve AI Agent Actions*.

## Related

> [card] [Automation rules](https://help.journeyfront.com/workflows/automation-rules.md)
>
> Fixed rules that act on every candidate the same way.

> [card] [Text messages](https://help.journeyfront.com/communication/text-messages.md)
>
> Phone numbers, opt-in and message templates.

> [card] [Add-ons and credits](https://help.journeyfront.com/account-setup/add-ons.md)
>
> Your AI credit balance and usage.

> [card] [Security roles](https://help.journeyfront.com/account-setup/security-roles.md)
>
> Give people the AI Agents permissions.


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