Responsible AI in background check review
Kizuna uses AI to help employers build policies, understand review patterns, and question assumptions behind their decisions. That assistance works alongside explicit policy rules, record expertise, and research-backed context. That keeps essential judgment with the employer, while supporting improved decision making over time.
Principles behind our approach
These principles guide where we use AI, the boundaries we set, and how we evaluate its role in background check review.
-
Match the technology to the task.
We choose where AI is used based on the specific task, the evidence that it helps, and the consequences of error. Some work benefits from language models; other work calls for explicit rules, established reference material, or human expertise.
-
Define the purpose. Test the output.
Each feature needs a defined purpose, appropriate data protections, and a process for checking its output. Testing, human review, and how uncertainty is handled should reflect the feature’s role and potential impact.
-
Question assumptions in past decisions.
Past policies and decisions can carry assumptions forward. We use research and ongoing evaluation to question those assumptions, while preserving room for individual circumstances and employer judgment.
Where does Kizuna use AI?
Use AI to draft and understand policies, explore review data, and investigate decision patterns. Your team reviews and approves policy changes before they take effect.
-
Build policies in plain language
Describe the role and your screening requirements in everyday language. AI helps translate that intent into a policy you can review and refine, with relevant research and established screening practices informing the starting point.
-
Understand the rules you’re applying
Ask questions about your policy, including its lookback periods, treatment of different dispositions, and role-specific criteria. Make the rules easier to understand before relying on them in a review.
-
Ask questions about your data
Explore review patterns through natural-language questions. Where are cases getting held up? Which criteria generate the most flags? Where might your team need a closer look?
-
Investigate patterns in decisions
Find recurring exceptions and differences across review teams. Explore whether they reflect job requirements, individual circumstances, or inconsistent policy application.
Should AI learn from your past hiring decisions?
Past decisions are useful evidence to examine. However, agreement with those decisions does not establish that the underlying criteria were justified. If an employer has always excluded theft convictions within seven years, repeating that pattern does not explain why seven years, or why the same restriction should apply to different roles.
Our approach brings those assumptions into the review. Relevant research can help employers examine how elapsed time, age, and conviction history affect the assessment, alongside the job requirements and the candidate’s circumstances.
Time Lens
Time Lens brings research-based context on elapsed time, age, and conviction history into individualized assessment. It helps reviewers examine the significance of a record rather than relying on its label alone. This research concerns subsequent convictions; it does not determine a person’s future job performance or make the hiring decision.
Time Lens →The research
A substantial body of peer-reviewed criminology and labor economics research examines how the likelihood of a subsequent conviction changes with elapsed time, age, and conviction history, and how criminal records affect employment. Our research library collects the studies that inform our approach, with their populations, methods, and limitations.
Research library →How is Kizuna different than a chatbot?
Natural-language assistance can make complex information easier to work with. A background check review also needs a reliable connection between the reported record, the employer’s criteria, the candidate’s circumstances, and the final decision. A detailed prompt alone does not establish those connections.
Review question
Standalone chatbot
kizuna What does the record mean?
Interpretation depends on supplied context
Classification with jurisdictional expertise
Which criteria apply?
Criteria supplied through prompts
Your defined criteria, applied consistently
What supports the assessment?
An explanation that requires checking
Research-based context with the record and role
What does the candidate add?
Candidate input must be supplied separately
Candidate Context in the review
How was the decision reached?
A chat transcript
A documented workflow to the decision
Standalone chatbot
What does the record mean?
Interpretation depends on supplied context
Which criteria apply?
Criteria supplied through prompts
What supports the assessment?
An explanation that requires checking
What does the candidate add?
Candidate input must be supplied separately
How was the decision reached?
A chat transcript
kizuna What does the record mean?
Classification with jurisdictional expertise
Which criteria apply?
Your defined criteria, applied consistently
What supports the assessment?
Research-based context with the record and role
What does the candidate add?
Candidate Context in the review
How was the decision reached?
A documented workflow to the decision
Data handling is a separate question. Where a background check report goes should follow the authorizations and agreements that cover it. Kizuna’s handling is described in our Privacy Policy, Data Processing Addendum, and subprocessors list. Our post on AI chatbots and background checks covers the employer’s side.
Built for accountability and oversight
Kizuna connects policy criteria, candidate context, and review documentation so your team can explain how decisions were reached.
Our AI Governance and Audit Support Package provides documentation to support your legal, compliance, and audit teams—from understanding how Kizuna works to preparing notices and assessments.
FAQs
Where does Kizuna use AI?+
We use AI to help employers build and understand screening policies, explore review data, and examine decision patterns. That includes questioning past practices, not simply learning to repeat them.
We use approved business services under appropriate data processing agreements, with controls over what information is shared, who can access it, how long it is retained, and whether it can be reused for model training. Personal information is limited to what the task requires.
AI supports this work alongside criminal-record expertise, policy rules, and research. The employer remains responsible for the hiring decision.
Does AI make the hiring decision?+
No. Kizuna produces a structured assessment: how the record was classified, which of the employer's criteria applied, the research-based context, and the candidate's response. The employer's reviewer reaches the decision and records the basis for it. Kizuna does not make or transmit employment decisions.
Can I just paste a background check into ChatGPT?+
A chatbot can explain unfamiliar language and help you identify questions to investigate. On its own it doesn't establish that the record was interpreted correctly, that the right criteria were applied, that the assessment rests on evidence, or that the candidate's circumstances were considered, and it leaves no documented basis for the decision. A background check report is also a consumer report full of a real person's personal information, so where it goes should follow the authorizations and agreements that cover it.
How do I know why a record was flagged?+
Kizuna shows which of your criteria the record met and the research-based context that applies to it, so a reviewer can see the basis for the flag and examine whether the concern is supported. The reviewer's assessment and decision are recorded in the review workflow.
What happens when a record is ambiguous?+
It goes to a person. Charges and dispositions that can't be interpreted with confidence from the report are routed to human review rather than resolved by guesswork, and the review isn't complete until they're resolved.
Do AI hiring laws apply to Kizuna?+
It depends on the jurisdiction, how you use Kizuna, and how much weight its output carries in your decision. Several jurisdictions regulate software whose output influences hiring decisions, and their requirements differ. Kizuna does not make that determination for you and does not audit your outcomes. What it provides is a structured review that connects the record, your criteria, research-based context, and the candidate's response, so you and your counsel can evaluate and explain how a decision was reached.
Is Kizuna a background check provider? What is its output?+
Kizuna is workflow software your team uses to review reports your screening provider furnishes. Kizuna does not compile or sell background reports, does not modify the report your provider delivered, and the provider's report remains the record of reference. Kizuna's output is your internal review record: the classification, the policy result, and the reviewer's decision. How that record is treated under federal and state consumer reporting law depends on how you and your provider use it. Our Platform Terms describe the responsibilities each party keeps.
How does Kizuna support individualized assessment?+
Flagged records carry the nature of the offense, the time elapsed, and the role in structured form, alongside the candidate's response and research-based context on how those factors relate to documented outcomes. Reviewers record their reasoning per case, so the assessment is documented, not just performed.
See it on your own policy and records
Bring your screening criteria and a few difficult records. We'll show how AI helps build and explain the policy, how the rules apply, and how research and candidate context come into the review.