Using the Workable Agent

The Workable Agent is an intelligent hiring assistant that helps you create jobs, define structured matching criteria and automate sourcing, screening and candidate engagement. It is designed to reduce manual work, improve consistency in evaluation and help you identify strong candidates faster. The Workable Agent includes:

• AI-powered sourcing & outreach: Agent finds and communicates with qualified candidates.
• Candidate evaluation: Agent instantly screens and enriches candidates.
• Smart pipeline management: Agent automatically advances candidates.
• Seamless handover: Get instant notifications when they’re ready.

The Job brief Agent is free — use it to create agentic jobs with a full set of matching criteria. The Sourcing, Screening, and Engagement Agents run on credits. One credit covers everything the Agent does to screen a single candidate: matching against your ideal candidate profile, outreach, and chat. Credits can be purchased via the Plan page.

Creating a job with the Workable Agent

To use the full Workable Agent capabilities, you must create an agentic job using the Job brief Agent.

1. Select Create new job with Agent from the Create new job button on the Jobs page or the homepage.
   
2. Follow the guided conversation to describe your hiring needs. It's advised not to copy/paste ready job descriptions during this process. You should provide the needed details gradually to generate the ideal candidate profile.

Creating an agentic job is only available when using the AI-powered job creation flow. This option is not available when creating a job from a new requisition.

The Agent will guide you through a structured chat to:

• Define the role and team context
• Capture required skills, qualifications and legal requirements
• Suggest relevant details based on your account’s historical data and similar roles
• Populate standard job editor fields such as location and workplace type
• Build a structured ideal candidate profile
• Generate a job description
• Generate targeted application form questions

Defining the ideal candidate profile

After collecting all necessary information (see example below), the Agent will shape its strategy to define the ideal candidate profile.

The ideal candidate profile is defined by the user during the job creation process as a set of matching criteria organised in 14 categories.

The Agent marks criteria as must-have based on its understanding, but users can manually add entirely new criteria, modify the existing ones, change visibility, or set disqualifying status. Candidates matching disqualifying criteria are marked differently, with automations possible on them. Each criterion can be configured as:

• Must have: If the candidate does not meet it, it will have a greater impact on the score calculation of not met criteria.
• Disqualifier: If the candidate does not meet a criterion marked as Disqualifier, it will be marked as disqualifying. The candidate will be disqualified only if this action is enabled in the Agent actions settings.
• Show in job post

Adding entirely new criteria or deleting existing ones is not supported after the job creation to ensure consistent and fair scoring across candidates.

Once you are satisfied with the list of matching criteria, the Agent will:

• Generate the job description and fill in any known job editor fields. You can freely modify the job description. Any changes to the job editor will not impact the matching criteria, and the Agent will not re-engage.

• Generate application form questions for must-have, public, and not easily evaluated by the CV matching criteria. You can freely modify the questions and mark any of them as disqualifying.

Jobs created with the Job brief Agent are tagged on the Jobs page and on the Job widget on the homepage with a relevant indication on the bottom right.

Upgrading existing jobs

Existing non-agentic jobs, or jobs created manually, can also be upgraded to agentic jobs via the job editor and the AI Agent tab by selecting the "Let's refine this role" option. 

When you select the upgrade option, the Agent reads the existing job editor content and:

• Keeps the company description and benefits unchanged.
• Expands qualifications and requirements, preserving the existing tone of voice.
• Generates a new job description and shows a side-by-side comparison with the current one so you can select the version that best fits your needs.
• Generates application form questions if the job has none; if questions already exist, no new ones are generated.

Using the AI Agent tab

Each agentic job includes an AI Agent tab in the job editor. This tab centralizes all AI-related settings and insights. From this tab, you can:

• 'Set up Agent actions' for the first time (available only after you publish an agentic job - even internally)
  
• View the Agent’s current status
• Start or stop Agent actions
• Review and manage the ideal candidate profile
• Configure sourcing, screening and engagement settings
• View performance insights

If the Agent is stopped, no new automated actions will be executed. Candidates already in progress (meaning they have already started the chat with the Agent) will not be interrupted, and scoring continues using the defined criteria.

Agent insights and performance metrics

The Insights section provides a summary of the Agent’s activity and key performance indicators. You can monitor metrics such as:

• Sourced: Passive candidates added by the Agent to the Sourced stage; their source will be set as "Workable Agent".

• Interested: Sourced candidates who successfully completed the chat.

• Evaluated: Candidates evaluated against the ideal candidate profile.

• Updated profile: Applicants who completed the chat and enriched their profile.

• Disqualified: Candidates disqualified due to unmet disqualifying criteria or lack of interest. Clicking the disqualified pill will direct you to a filtered view of the job pipeline that shows only those candidates disqualified by the Agent.

• Handovers: Candidates fully screened and handed over to a user. All non-disqualified candidates are, in principle, handed over to a human, in the sense that, going forward, it's up to humans to decide what to do with the candidate. The agent also mentions the user in the following scenarios: 

  • Auto-advance/move is ON, and a candidate has passed all the evaluation thresholds.

  • During chat, a candidate has explicitly requested for a human.

• Advanced: Candidates moved to a later stage based on score thresholds.

Clicking a metric redirects you to the Candidates page with the relevant filters applied. 

Stop Agent

Agent’s status is displayed and you can stop the Agent if needed from the relevant action button. Stopping implies that the below Agent automated operations are halted and no further actions are taken by the Agent on new and existing candidates:

• Auto candidate sourcing of passive candidate
• Communication with candidates
• Auto advance and auto disqualify

Also, to maintain a proper candidate experience, candidates that are in flight of Agent’s actions will not be impacted. For example, a candidate receives an email to join the chat session. At 10.00 the candidate enters the chat. At 10.01 you stop the Agent. This candidate is already in the chat (in-flight) and thus will not be impacted.

Ideal candidate profile

The Screening Agent evaluates every candidate against the ideal candidate profile and assigns a normalized score out of 100. Must-have criteria are weighted more heavily. The evaluation framework considers:

• Completeness of the candidate profile, including CV and application answers

• Validity and consistency of provided information

• Complexity of defined criteria

• Alignment between candidate data and defined requirements

• Subjective elements such as transferable skills

Editing or creating a new matching criterion is not possible during this release, but will be available shortly after. However, you can modify whether a criterion:

• Is Must have: Upon change the scores of all candidates is reevaluated
• Is Disqualifying: Upon change impacted candidates are denoted

After a change, new and in-flight candidates are evaluated based on the new settings.

Agent actions settings

For agentic jobs, you can enable the following actions if you have credits available.

Any manual user activity on a candidate immediately stops the Agent from taking further action. For example, if you leave a comment, the Agent will no longer move, disqualify, or request additional information, as it assumes a human has taken over. Agent actions will be initiated for applied and sourced candidates via the Agent or via Search with AI (manually by the user but not past candidates) or via the Workable extension.

The Agent also does not reprocess candidates it has already completed. If the criteria are later updated and a previously reviewed candidate now qualifies, the Agent will not automatically move them. Instead, users will be notified about affected candidates and can manually take bulk actions.

Stop Agent actions

You can stop Agent actions on a candidate basis from the following locations:

• Under Communication > Messages tab.
  
• From the toolbar actions under the three dots menu.
  

Screening Agent - matching against ideal candidate profile

The Agent’s evaluation framework considers the following: 

• The completeness of candidate input (CV, ATS profile, application form questions and candidate custom fields)
• The validity of data (incomplete parsing,  conflicting information, etc.)
• The complexity of the user defined matching criteria (criterion A and criterion B or criterion C, where A, B, C are all different types of data, such as numeric, categorical, hierarchical, etc.)
• The misalignment of data formats between candidate-provided and user-provided text data
• The subjectiveness of the evaluation process (e.g. transferable skills, knowledge domain mappings, synonyms etc.) 

The Agent's output comprises two core sections: an evaluation summary and a detailed match against the ideal candidate profile. Note that the score on the candidate avatar in the candidate list appears only in the All, Sourced, and Applied stages. For the remaining stages, you will see only the evaluation summary, and the avatar won't display the score.

Evaluation summary

This section provides:

• The candidate score: The score as provided by the Agent is aggregated and normalized to 100%. The scoring algorithm scores must haves with twice the marks as non-must-haves. More specifically:
  • Nice to have = 1 point
  • Must have = 2 points
  • Partially met: half the points
• Missing must-haves: The number of missing must haves (if any).
• Summary: A high level text summary of the candidate’s evaluation. 
• Status: Candidates that are still under evaluation by the Agent will be notated as “Reviewing..”.

Detailed evaluation

This section provides the candidate’s fit to the ideal candidate profile as evaluated by the Screening Agent. 
The matching criteria are grouped as Must-have and Nice-to-have. Each criterion has one of the following states:

Status	Description
Met	The candidate fully matches the requirement based on available evidence.
Partially met	The candidate partially satisfies the requirement.
Not met	Evidence exists but does not meet the requirement.
Disqualifying	A requirement marked as disqualifying that was not met.
Unknown	There is insufficient information to evaluate the requirement.
Reviewing	The Agent is still processing and evaluating the candidate.

All matching outcomes are also supported by relevant reasoning supporting the Agent’s claims. 

If needed, you can manually modify any matches and update their states by selecting "Edit candidate's fit". 

Note that updating a candidate's matches will stop the Agent from further engaging with this candidate, i.e., send any message or advance them to the next stage. Chat links will remain active, but the candidate's responses will not be evaluated, as those might override your choices.

Sorting candidates by highest score

For agentic jobs (newly created or upgraded to agentic), the default sorting will be by the highest score to see ready-to-review candidates first. In the Sourced & Applied stages, under the 'Sort & filter' menu, you can view all available options.

Sourcing Agent - sourcing matching passive candidates

The Sourcing Agent aims to identify and source qualified, interested candidates for a job. Using Search with AI suggestions, it targets passive candidates who match the role.

First, it retrieves the top candidates suggested by Search with AI. The Screening Agent then reviews them and filters out anyone who clearly does not meet disqualifying criteria. The remaining candidates are added to the Sourced stage. If some are disqualified, the Sourcing Agent will automatically source additional candidates until the target is reached.

For jobs with multiple locations, the best profiles are selected across all locations.

For candidates added to the Sourced stage either by the AI agent or via Search with AI, any message templates, video interviews, assessments, or self-schedule automations set on that stage will be skipped. Automations in the applied stages will be triggered normally. Note that the "thank you" automation will not stop the Agent from contacting the candidate, but you might want to reconsider them, as the candidate will get three emails upon application (the workable email with their application info, the thank you automation and the Agent email.)
All other automations (in the applied or following stages) will be triggered as configured, but those will result in stoping the Agent from taking any other action and invalidating the chat links.

Engagement Agent - Reach out to candidates

The Engagement Agent is responsible for: 

• Communicating with the candidates, inviting them to a chat session in order to gather interest and/or enrich their profiles
• Autonomously chatting with candidates
• Acting on the candidates (move, disqualify, revert, comment etc.) as an outcome of the chat

Candidate communication

If the "Contact sourced candidates" is enabled on Agent actions settings, the Agent will send personalized emails to sourced candidates via the Agent or via Search with AI (manually by the user but not past candidates) or via the Workable extension to invite them to a chat session in order to check their interest and complete their profile.

Other sourced candidates, for example, those manually uploaded or via RSC, will not be contacted, as prior communication external to the system may already exist.

Similarly, the Agent will email candidates who applied to clarify any unknown matching criteria as specified in the Agent settings. 

In both cases, emails are sent from a system-generated no-reply address and are signed by the selected job collaborator in the Agent settings. The tone of voice, included information and number of follow-ups are defined in the AI Agent settings as well.

Chat with candidates

The chat is a guided conversation between the candidate and the AI Agent. Its purpose is to:

• Confirm interest in the role 
• Request a CV if missing 
• Clarify unknown or incomplete matching criteria 
• Enrich the candidate profile

At the bottom of the chat window candidates can access your Privacy notice (if one is uploaded under Settings > Compliance).

If a candidate uploads a CV during the chat, the Agent parses the document and updates the evaluation against the ideal candidate profile. The Agent may continue asking follow-up questions until all required information is collected.

Before the conversation ends, the Agent prompts the candidate to share any questions they would like passed to the hiring team.

Chat termination/expiration and completion statuses

The chat invitation is valid for 6 days. If a candidate does not open the chat within 6 days, the link will expire regardless of the number of follow-up emails sent. Once the chat is opened and initiated, the candidate will have 7 days to complete the conversation. The timeline is not extended. The remaining time will be displayed on the footer.

Candidates can continue the chat any time within those 7 days. The conversation will continue from the point it was previously stopped. If the candidate has not completed the chat depending on the number of allowed follow-ups defined in Agent settings, the Agent will send reminders every 2 days, asking the candidates to enter the chat.

The chat is completed for one of the following reasons:

• completed_successfully: the candidate answered all Agent questions
• not_interested: the candidate mentioned that they are not interested
• expired: The candidate started the chat, but never completed it and the link has expired.
• irrelevant_responses: The candidate did not provide relevant responses to questions. The Agent may terminate the chat early if need be.
• human_handover: the candidate asked for a human to step in
• other: any other situation

Once the chat is terminated (for any reason) the Agent will leave a comment to the candidate’s profile and include any relevant information as well as the complete chat conversation log. 

After the chat with a passive candidate, if the Agent disqualifies them (due to a met disqualifying criterion) while on the Sourced stage, any disqualification automations (if any) will be skipped.

 

Agent retrospective logic

Changes to Agent settings apply not only to new candidates but also retroactively to existing ones. Specifically, non-disqualified candidates created in the last 7 days, where no user activity has stopped the Agent, are reprocessed in the following scenarios:

• The Agent is re-activated on an agentic job
• Agent settings are modified
• A non-agentic job is transformed to an agentic one and the Agent is enabled
• AI features are enabled at the account level

Candidates queued for reprocessing are indicated on the candidate profile. 

When specific Agent settings are toggled on:

• Source passive candidates: The auto-source algorithm re-initiates and new matching passive candidates are added to the job.
• Contact sourced candidates: Agent-sourced or Search with AI candidates on the Sourced stage, created within the last 7 days, are contacted; follow-ups are sent on a 2-day schedule.
• Request missing details: Non-disqualified candidates who applied in the last 7 days and are still on the Applied stage are contacted; follow-ups are sent on a 2-day schedule.
• Auto-advance candidates: Eligible candidates from the last 7 days meeting the score threshold are advanced; active chat links continue but no further move or disqualification actions are taken by the Agent.
• Auto-disqualify candidates: Eligible candidates from the last 7 days not meeting a disqualifying criterion are disqualified; active chat links continue but no further actions are taken.

When specific Agent settings are toggled off:

• Contact sourced candidates: New sourced candidates will not be contacted; existing ones will not receive follow-up emails.
• Request missing details: New applicants will not be contacted; existing ones will not receive follow-up emails.
• Auto-advance candidates: Candidates will no longer be automatically advanced.
• Auto-disqualify candidates: Candidates will no longer be automatically disqualified.

 

Fair use and employer responsibility

The matching criteria used by the Workable Agent are defined by the employer during the agentic job creation.

The Agent is designed to support consistent application of those criteria. Employers remain responsible for reviewing candidates and ensuring their hiring practices comply with applicable employment and anti-discrimination laws.
 

Enabling or disabling AI features

Admin users can control AI functionality for the entire account from the Workflow settings. Disabling AI features turns off all features used for job creation, sourcing, screening, and ranking across all jobs.

FAQs

Why am I no longer seeing match scores for new candidates?

Match scores were generated by the Screening Assistant, which is being replaced by the Workable Agent as part of a gradual rollout. Once the transition is complete, scores will no longer be generated for new candidates on jobs that haven't been converted to agentic.

To continue scoring and ranking candidates, you'll need to upgrade your job(s) to use the Workable Agent. Once upgraded, the Agent will evaluate candidates against your ideal candidate profile and rank them accordingly.

Where did the "Best match" filter go?

The Best match filter has been removed. For jobs using the Workable Agent, candidates are ranked by their match score by default — so the highest-scoring candidates will appear at the top of your pipeline without needing to apply a filter.

For jobs that haven't been upgraded to the Agent yet, candidate ranking is no longer available. To restore ranking for those jobs, convert them to agentic.

Does Workable Agent support languages other than English?

The Workable Agent doesn't support languages other than English yet.

Will Workable Agent run for all my agentic jobs once I have credits?

Once your account has credits, the Agent is available across all your agentic jobs.

However, candidates created more than 7 days before the Agent was enabled will not be automatically screened. The Agent will evaluate new candidates moving forward. If credits are available, you can manually initiate Agent screening for older candidates from the candidate profile.

Will Workable Agent engage with candidates in all stages?

No, Workable Agent will engage with candidates only during the Applied and Sourced stages.

How can I see which candidates the Agent disqualified?

Candidates disqualified by the Agent will have relevant items added to their profile timeline, explaining the disqualification action based on the candidate's profile information.

On the job editor's AI Agent tab, you can view the number of candidates disqualified by the Agent. Clicking the disqualified pill will direct you to a filtered view of the job pipeline that shows only those candidates disqualified by the Agent.

Can I edit the ideal candidate profile after a job is published?

Currently, matching criteria cannot be added or removed from the ideal candidate profile once a job is published, as this would lead to inconsistent scoring across candidates. If you need to make significant changes to the ideal candidate profile, we recommend creating a new agentic job. This limitation will be addressed in a future release.

Can a recruiter oversee the Agent's actions?

Yes. A recruiter can review, edit, and override actions, decisions, or recommendations made by the Agent at any point. More specifically:

• The recruiter can manually modify any matches and update their states by selecting "Edit candidate's fit".
  
• If needed, you can revert any candidates the Agent disqualified.
• If the email asking the candidate to chat with the Agent has already sent, a recruiter cannot take it back, but can "stop the Agent" making this link invalid, and then manually follow up with them. You can stop the Agent:
  • On the candidate level under the Communication > Messages tab.
    
  • On a job level under the AI Agent tab on the job editor.
    
  • On an account level by disabling AI feature for the entire account from the Workflow settings.

How can I check which jobs the agent is activated for?

You can see it from the 'Agent is active' mention in the Jobs page. Note that the Agent can be enabled only for agentic jobs.

Why does creating a job with the Workable Agent take longer than a standard job?

Creating an agentic job goes beyond writing a standard job listing. When hiring, you evaluate candidates on more than just their experience and education — there are often other characteristics you'd typically explore in a first screening call that never make it into a job ad. The Job brief Agent captures all of this upfront: it helps you build a richer, more precise job description while also defining the full set of criteria you actually use to assess candidates.

The more specific you are during this process, the more accurately the Agent can score and evaluate candidates — so the time invested at this stage directly improves the quality of your results.

If you're hiring for a similar role in the future, you can duplicate an existing agentic job to reuse the ideal candidate profile and save time on setup.