TL;DR:
AI resume screening reduces manual review work, but only delivers fair, defensible results when paired with human oversight.
Ashby, BambooHR, and Greenhouse each takes a different approach to applicant tracking, and the right fit depends on your team's size, hiring volume, and process requirements.
Composio connects to Ashby, BambooHR and many other tools through a SOC 2-certified integration layer, so your recruiting team can automate the screening workflow without building integration code from scratch.
Applicant tracking systems like Ashby, BambooHR, and Greenhouse help recruiting teams manage job openings, applications, interviews, candidate communication, and hiring decisions. Each platform solves a slightly different problem.
Ashby is designed for recruiting teams that need advanced analytics, sourcing, scheduling, and applicant tracking in one system. BambooHR connects recruiting with wider HR processes like employee records and onboarding. Greenhouse focuses heavily on structured hiring, interview plans, and consistent scorecards.
These platforms already reduce a large amount of manual recruiting work. However, recruiters still spend considerable time searching for information, summarizing candidate records, checking whether feedback has been submitted, preparing hiring reports, drafting messages, and moving information between systems. This is where an AI agent can help.
When an AI agent is securely connected to an applicant tracking system, it can retrieve approved information, summarize records, prepare reports, identify incomplete tasks, draft communications, and complete permitted administrative actions. Recruiters can interact with the system through natural-language instructions rather than manually opening and reviewing every candidate profile.
This guide explains how Ashby, BambooHR, and Greenhouse work, where each platform is strongest, and how connecting them to an AI agent can improve day-to-day recruiting operations without handing important hiring decisions over to automation.
What AI-assisted recruiting looks like in practice
AI recruiting does not need to mean automatically ranking and rejecting every applicant. A safer workflow uses AI to support specific, reviewable steps:
An application enters the ATS.
Candidate information is parsed into structured fields.
The AI agent retrieves the relevant information for the recruiter.
The candidate is compared with documented, job-related requirements.
The AI agent prepares a factual summary showing which requirements have and have not been evidenced.
A human recruiter reviews the original application and the summary.
The recruiter decides whether the candidate advances.
The decision and its reason are recorded in the ATS.
The recruiting team periodically audits outcomes for possible adverse impact.
AI-assisted recruiting tools
1. Ashby
Ashby is an all-in-one recruiting platform designed to bring several parts of the hiring process into one system. Its platform includes:
Applicant tracking
Candidate sourcing
Recruiting CRM
Interview scheduling
Hiring analytics
Job posting
Offer management
Recruiting automation
Instead of using one platform for applicant tracking, another for scheduling, and another for reporting, a company can use Ashby to manage the full recruiting process.
Ashby is particularly useful for growing companies with more complicated hiring operations. Recruiting teams can create different interview processes for different roles, automate routine steps, monitor candidate movement, and analyze hiring performance from a central platform.
How Ashby works
A typical Ashby workflow looks like this:
A recruiter creates a job and defines its interview process.
The role is published to the company’s careers page and selected job boards.
Applications enter Ashby as candidate profiles.
Recruiters review applications and move candidates through defined stages.
Interviewers submit structured feedback and scorecards.
The recruiting team schedules interviews, manages communication, and prepares offers.
Hiring leaders use Ashby’s analytics to review conversion rates, hiring speed, sourcing performance, and pipeline health.
Because candidate data, interview feedback, and hiring reports sit in the same platform, Ashby can become a useful source of context for an AI assistant.
How connecting an AI agent to Ashby can help
Connecting Ashby to an AI agent gives recruiters a faster way to review incoming applications against the job requirements already defined in the platform. Instead of opening each profile individually, the recruiting team can ask the agent to retrieve application data, compare it with documented criteria, summarize evidence, and flag gaps, all for recruiter review before any candidate advances.
Depending on the permissions granted, the agent may also be able to complete approved actions inside Ashby, while recruiters remain responsible for reviewing outputs and making advancement decisions.
Here are some screening-specific tasks recruiters can ask an AI agent to help with:
“List all new applications received for the senior product manager role in the past 48 hours.”
“For each new applicant, state whether their resume provides clear evidence, partial evidence, or no evidence for each of the following requirements: [list criteria]. Quote or reference the relevant experience where it exists.”
“Flag any applications where key required fields, such as years of relevant experience or right-to-work information, are missing or unclear.”
“Summarize the qualifications of the applicants currently in the initial review stage without ranking them or making advancement recommendations.”
“Identify applications where the candidate’s stated experience does not match the minimum requirements as defined in the job spec. List each gap specifically.”
“Prepare a structured summary of this applicant’s resume for recruiter review. Do not infer characteristics, fill in gaps, or make a hiring recommendation.”
“Show me which applications for the account executive role have been in the initial review stage for more than three days without a recruiter decision recorded.”
“Compare the stated qualifications of these five applicants against our essential requirements. Present the evidence side by side. Do not score or rank.”
2. BambooHR
BambooHR is a broader human resources platform that includes applicant tracking alongside employee records, onboarding, reporting, time-off management, performance tools, and other HR functions.
Its applicant tracking system helps internal HR teams:
Create and publish job openings
Collect applications
Store resumes and candidate information
Move candidates through hiring stages
Coordinate hiring-manager feedback
Communicate with applicants
Prepare offer information
Transfer successful candidates into employee onboarding
BambooHR can be a good fit for small and midsized organizations that want recruiting and post-hire HR administration to remain closely connected. Its ATS is integrated with the wider BambooHR platform, reducing the need to move information manually from a recruiting system into a separate employee database.
How BambooHR works
A hiring team can create a vacancy inside BambooHR and publish it to a careers page or supported job boards. Candidate applications are collected in a central hiring pipeline, where recruiters and managers can review resumes, leave comments, update statuses, and coordinate interviews.
When a candidate is hired, their information can flow into the company’s HR records and onboarding process. This makes BambooHR useful for companies that want one connected workflow from application through employment.
How connecting an AI agent to BambooHR can help
Connecting BambooHR to an AI agent can make it easier to review incoming applications against the role requirements stored in the platform. The agent can retrieve applicant information, compare it with documented criteria, summarize what each application does and does not evidence, and prepare that output for a recruiter to review. Because BambooHR may also contain sensitive employee information, the connection should be scoped to applicant tracking records only.
For example, ask the AI agent to:
“List all applications received for the customer success manager role, grouped by current review status.”
“For each application currently in the initial review stage, summarize what the resume states about the following requirements: [list criteria]. Do not infer or add information not present in the application.”
“Flag applications where the candidate has not provided evidence for one or more essential requirements. List each missing criterion specifically.”
“Identify any applications where the resume is incomplete, illegible, or missing a key section like work history or contact details.”
“Prepare a side-by-side summary of these three applicants against our stated essential requirements. Present evidence only. Do not rank or recommend.”
“Show me applications for the operations role where the candidate’s experience appears potentially relevant but uses different job titles or terminology than our job description.”
“Summarize this applicant’s resume in plain language, separating clearly evidenced experience from experience that is unclear or only partially described.”
“List applications that have been waiting for an initial recruiter review for more than three business days.”
“Prepare an onboarding guide for the candidate we have marked as hired.”
3. Greenhouse
Greenhouse is a recruiting and hiring platform best known for its structured hiring approach.
Structured hiring means that the company defines the requirements of a role, interview stages, interviewer responsibilities, and evaluation criteria before candidates begin moving through the process. Interviewers then assess candidates using consistent scorecards instead of relying entirely on informal impressions.
Greenhouse supports:
Applicant tracking
Candidate sourcing
Interview planning
Interview scheduling
Structured scorecards
Recruiting reporting
Candidate communication
Offer management
Onboarding
Recruiting automation and AI-assisted features
Greenhouse also provides AI features for tasks like summarizing scorecard feedback, supporting sourcing communication, and organizing recruiting information.
How Greenhouse works
A recruiting team starts by defining a role and building an interview plan. Each interview stage is assigned a purpose, and interviewers receive scorecards tied to the competencies they are expected to evaluate.
Candidates then enter the pipeline through applications, referrals, sourcing, or agency submissions. Recruiters move candidates between stages, schedule interviews, collect scorecards, and monitor whether the process is moving consistently.
This structure makes Greenhouse particularly useful for teams that want a documented and repeatable hiring process.
How connecting an AI agent to Greenhouse can help
Connecting Greenhouse to an AI agent can help recruiting teams apply the competencies and criteria already defined in the platform to incoming applications more consistently. The agent can retrieve application data, compare it against the structured criteria in the interview plan, summarize what each application does and does not evidence for each competency, and prepare that output for recruiter review, without making advancement decisions.
For example, ask an AI agent to:
“List all applications currently in the initial application review stage for the enterprise account executive role.”
“For each applicant in the initial review stage, state whether their resume provides clear evidence, partial evidence, or no evidence for each competency defined in the interview plan. Reference the relevant experience directly.”
“Do not produce a hiring recommendation. Present the evidence recorded in each application for each required competency.”
“Flag applications where the candidate has provided no evidence for one or more essential competencies as defined in our scorecard.”
“Identify applications where the candidate’s experience appears relevant but uses terminology different from our job description. List the specific roles and descriptions involved.”
“Summarize this applicant’s background against our defined competencies. Separate explicitly stated experience from experience that is implied or unclear. Do not infer characteristics such as age, gender, or background.”
“Compare these four applicants against the competencies defined for this role. Present results as a structured table showing evidence status per competency. Do not rank.”
“Show me applications that have been in the initial review stage for more than four days without a recruiter decision.”
“Prepare a structured screening summary for this application that a recruiter can use as a starting point for their review. Flag any areas where the information is ambiguous or missing.”
The agent can help apply Greenhouse's existing competency framework to incoming applications more consistently. The hiring team still needs to define what each competency means, set the evidence standard, and decide whether each candidate advances based on what their application actually contains.
Which recruiting platform is right for your team?
The three platforms overlap, but they are not identical.
Platform | Particularly useful for | Main strength |
|---|---|---|
Ashby | Scaling recruiting teams with complex operations | Combines ATS, sourcing, scheduling, CRM, automation, and analytics |
BambooHR | Small and midsized companies wanting recruiting and HR in one system | Connects applicant tracking with employee records and onboarding |
Greenhouse | Teams that prioritize structured interviews and consistent evaluation | Detailed interview plans, scorecards, workflows, and recruiting controls |
The best choice depends on your hiring volume, organizational size, reporting needs, HR technology stack, and the level of process control you require.
An AI agent connected through Composio does not replace any of these platforms. The ATS remains the official system of record. The agent provides a conversational way to retrieve information, prepare work, and complete permitted actions across the connected system.
Example prompt for responsible candidate summaries
A prompt such as “Choose the ten best candidates” gives the AI too much discretion and does not define what “best” means.
Use a prompt with explicit boundaries instead:
Review the candidates for the account executive role using only the approved criteria listed below. For each criterion, state whether the resume provides clear evidence, partial evidence, no evidence, or insufficient information. Quote or reference the relevant experience where possible. Do not infer age, gender, ethnicity, disability, family status, socioeconomic background, or personality. Do not rank candidates or make the final advancement decision. Present the results for recruiter review.
This produces a more useful and auditable output than asking the system for an unexplained score.
How to use recruiting platforms fairly
AI can make recruiting faster, but speed does not automatically produce fairer decisions. A faster version of a poorly designed process will reproduce its existing problems more efficiently.
The following safeguards should apply whether you use Ashby, BambooHR, Greenhouse, or another ATS.
1. Define the criteria before reviewing candidates
Agree on the requirements before applications are opened.
Separate them into:
Essential requirements
Preferred qualifications
Skills that can be learned after hiring
Genuine disqualifying conditions
Each criterion should have a direct relationship to the work.
For example, “has managed enterprise software implementations involving at least three stakeholder groups” is more useful than “comes from a prestigious technology company.”
2. Focus on competencies and evidence
Avoid using university names, previous employers, personal interests, home addresses, names, or career-path conventions as shortcuts for quality.
Assess evidence such as:
Work completed
Problems solved
Relevant responsibilities
Results achieved
Tools used
Scope of ownership
Demonstrated knowledge
Relevant certifications where genuinely necessary
Be careful with requirements like a fixed number of years of experience. Someone with four highly relevant years may be more qualified than someone with ten loosely related years.
3. Do not let AI make the final hiring decision
AI can prepare a summary, organize evidence, or highlight missing information.
A qualified person should decide whether a candidate advances, is rejected, receives an offer, or is removed from consideration.
Human review should be real, not a rubber stamp. Clicking “approve” without reading the application, source evidence, and AI output is not effective oversight.
4. Do not ask the AI to infer sensitive characteristics
Do not ask the AI agent to infer or evaluate:
Race or ethnicity
Religion
Disability or health status
Sexual orientation
Gender identity
Pregnancy or family plans
Political beliefs
Union membership
Socioeconomic class
Nationality beyond lawful work-authorization requirements
You should also avoid indirect substitutes, like evaluating a person based on their postcode, graduation year, photograph, name, or unexplained employment gaps.
5. Keep sensitive information out of prompts
Only provide the information needed for the task.
A candidate-summary task may need the resume, job requirements, and relevant application answers. It generally does not need unrelated background-check data, medical information, demographic monitoring data, compensation history, or internal employee records.
Do not paste candidate data into an unapproved personal AI account.
6. Use the minimum required permissions
Connecting an ATS does not mean the AI agent needs permission to access or change everything inside it.
A useful permissions structure could be:
Read-only access for reporting and summaries
Limited write access for recruiter notes or draft records
Approval required before candidate-stage changes
Approval required before sending messages
Approval required before publishing jobs
No permission to delete candidate records
No access to unrelated employee information
Review connected applications periodically and remove access that is no longer needed.
7. Require approval for external or consequential actions
The AI agent can prepare a task without immediately carrying it out.
Require confirmation before it:
Publishes a job
Sends candidate communication
Moves a candidate to a different stage
Rejects an application
Creates or changes an offer
Deletes or exports candidate information
Shares candidate data with another application
This reduces the risk of a misunderstood instruction causing a real-world hiring error.
8. Verify every candidate summary against the source
AI summaries can omit context, overstate experience, combine unrelated facts, or interpret unclear language too confidently.
Recruiters should be able to trace every important statement back to the candidate’s application, resume, assessment, or interview feedback.
Ask the agent to distinguish among:
Explicitly stated
Reasonably supported
Not stated
Unclear
Requires recruiter verification
“Not mentioned in the resume” should not automatically be treated as “the candidate cannot do this.”
9. Review rejected applicants, not only shortlisted applicants
Bias can remain hidden when a team only reviews the candidates selected by the system.
Periodically take a sample of candidates who were not advanced and have a recruiter independently review them. Look for:
Transferable experience that the criteria missed
Alternative job titles
Nontraditional career paths
Career breaks
International qualifications
Equivalent skills described using different terminology
Relevant experience outside conventional employment
This is one of the most practical ways to find false negatives.
10. Monitor results across the hiring pipeline
Track outcomes at each stage:
Applied
Passed initial review
Recruiter screen
Hiring-manager interview
Assessment
Final interview
Offer
Hire
Look for large differences in selection rates between groups where collecting and analyzing this information is lawful.
A difference does not automatically prove discrimination, but it should prompt an investigation into the criteria, data, process, and decision-making behavior.
11. Give candidates appropriate notice
Be transparent about how AI is used.
Candidate-facing information should explain:
That AI supports parts of the hiring process
What kind of information is processed
The purpose of the processing
Whether a person reviews the output
How long the data is retained
Which service providers receive it
How candidates can ask questions
Whether a manual review or accommodation process is available
Legal requirements vary by location, so involve privacy, HR, and legal specialists before deploying automated hiring tools across different jurisdictions.
12. Keep an audit trail
Record:
The job criteria used
The prompt or workflow version
The data sources accessed
The AI output
The person who reviewed it
The final decision
The documented reason for the decision
Any later correction
This allows the organization to investigate errors, respond to candidate questions, and determine whether a problem came from the criteria, source data, integration, AI output, or human reviewer.
You can start with Composio's free tier, connect your Ashby or Lever account, and have candidate notifications routing to your Slack channel before your next stand-up. No credit card required.
FAQs
What can AI not assess in a resume?
AI cannot reliably evaluate soft skills, cultural fit, or interpersonal communication style. While AI tools can analyze behavioral language patterns with some accuracy, they cannot observe how a candidate interacts with a team, handle subtle problem-solving signals that fall outside their scoring framework, or predict how someone will fit into a specific team dynamic. Human judgment and structured interviews are required for these dimensions.
How do I explain AI screening to candidates?
Update your job application privacy policy to state that AI is used as an initial screening assistant, and provide a clear opt-out path for candidates who prefer manual review. This is a candidate trust signal that positions your organization as transparent. Legal requirements vary by jurisdiction, so involve your privacy and legal team before deploying automated screening tools.
Does AI resume screening work with my existing ATS?
Yes, if you use a tool like Composio as your integration layer. Composio supports pre-built connections for Ashby, BambooHR, Lever, Slack, Notion, and Google Sheets, so your AI screening workflow connects to your existing stack without custom coding or ongoing API maintenance.
Is candidate data safe when using AI screening tools?
It depends on which vendor you choose. Look for minimal data retention policies (clear limits on how long candidate data is stored after processing), SOC 2 certification, and end-to-end encryption. Composio's integration layer is SOC 2 and ISO 27001 certified with data encrypted in transit and at rest. For strict zero-retention requirements, evaluate whether the vendor stores data at rest or passes it through without storing it.
Glossary
Applicant Tracking System (ATS): Software that manages job postings, candidate applications, interview stages, and hiring decisions in one place. Ashby, BambooHR, and Greenhouse are all examples of ATS platforms.
Adverse impact: A measurable difference in selection rates between groups of candidates at a given stage of the hiring process. A disparity does not automatically indicate discrimination, but it should prompt a review of criteria, data, and decision-making behavior.
Audit trail: A recorded log of every action taken during a hiring workflow, including the criteria used, the AI output, the reviewer, the final decision, and the documented reason. Enables investigation of errors and responses to candidate questions.
Competency: A specific, job-related skill or behavior used to evaluate candidates consistently. Defining competencies before reviewing applications reduces reliance on informal impressions.
Integration layer: Software that connects two or more platforms so they can exchange data securely without requiring custom code for each connection. Composio acts as an integration layer between ATS platforms, AI tools, and communication channels like Slack.
ISO 27001: An international standard for information security management. Certification confirms that a vendor has documented controls in place for managing risks to data confidentiality, integrity, and availability.
SOC 2 certification: An audit standard for service organizations, assessing controls related to security, availability, processing integrity, confidentiality, and privacy. SOC 2 Type II certification covers a period of ongoing operation, not just a point-in-time assessment.
Structured hiring: A recruiting approach in which role requirements, interview stages, interviewer responsibilities, and evaluation criteria are defined before candidates enter the process. All candidates are assessed against the same criteria using the same scorecard format.