AIM in practice · Fictional teaching example
An AI hiring decision, worked through with AIM.
An AI hiring tool should be evaluated by the decision it changes, the people affected, and the safeguards the organization can maintain. This example uses AIM to turn an attractive efficiency proposal into a documented leadership judgment.
The proposal: let AI choose the shortlist
A financial institution wants to reduce the time its recruiting team spends reviewing applications for customer operations roles. A vendor proposes ranking applicants from their résumés and advancing only the highest-ranked group to human review. Everyone outside that group would receive an automatic rejection.
The vendor describes its product as decision support because people conduct the interviews. The relevant decision happens earlier: who gets considered for an interview? In this proposed workflow, the system determines who is excluded. Under AIM’s role distinction, that step is decision automation, even though a person makes the final hiring choice.
The team has a product demonstration and a general accuracy claim. It does not yet have evidence tied to the role, an accessible alternative process, a useful explanation of individual exclusions, or a working challenge process.
Gate 0: can the institution own this decision?
Gate 0 asks whether this defined use should be permitted before procurement or deployment. The review records what is known, what is missing, and who must resolve each gap.
- Role definition
- The initial proposal obscures an automated exclusion behind a later human interview. The recruiting owner must document the full workflow, including default outcomes and override authority.
- Baseline duties and data stewardship
- The team cannot yet demonstrate relevant data use, applicant notice, accessibility, or protection of sensitive information. Privacy, accessibility, and recruiting owners must specify and review those arrangements.
- Inequality and lifecycle effects
- A general vendor claim does not show how this workflow treats applicants for these roles. The institution needs an appropriate evaluation plan, its limitations, and monitoring arrangements before relying on the ranking.
- Accountability and moral agency
- No one has been assigned to explain an exclusion, reconsider it, or stop automated screening. A named recruiting decision owner must have actual authority to act.
- Organizational and vendor readiness
- The vendor can change its scoring model without a defined institutional review. Procurement and the technical owner must establish change notice, reviewable records, and workable suspension and exit arrangements.
These are unresolved conditions of use. Completing a checklist or finding a time saving does not resolve them.
Use Rights, Reasons, and Results to examine the workflow
For the applicant, examine whether the process preserves accessible treatment, intelligible notice, and a way to challenge an exclusion. For the recruiter, examine whether staff can understand and question the recommendation and have time and authority to exercise judgment. For the institution and community, examine whether responsibilities, evidence, and the effects of screening can be reviewed over time.
The Triangle applies Rights, Reasons, and Results across each relationship. A final human signature would not, by itself, repair a process that prevented an applicant from reaching human consideration.
The Orbit adds wider questions: what recruiting capability could be lost through dependence on the vendor, how would hiring continue during an outage, and what effects would repeated exclusions have on trust and access to opportunity?
The judgment: redesign the proposed use
Decision: redesign. The institution does not authorize the proposed ranking and automatic-rejection workflow. In this scenario, the missing explanation, recourse, evidence, and stop authority are substantial enough that the proposal must change before reconsideration.
The team proposes a narrower administrative use: formatting application materials and coordinating interview scheduling. Recruiters would review every application against the documented role criteria; the tool would neither rank applicants nor exclude them. The revised scope requires a fresh review of privacy, accuracy, accessibility, and operational responsibilities before any pilot.
A later proposal to support or automate selection would receive its own Gate 0 review. Approval of an administrative use would not extend to ranking, new roles, new data uses, or a materially changed vendor system.
A completed decision memo for this scenario
- Decision and scope
- Redesign the customer operations screening proposal. Automated ranking and rejection are not authorized. Administrative assistance is a separate proposal awaiting review.
- People affected and AI role
- Applicants, recruiters, hiring managers, and the institution-community. The original shortlist gate would automate an exclusion decision.
- Evidence reviewed and missing
- Reviewed: proposed workflow, vendor demonstration, and product claims. Missing: role-relevant evaluation, accessible alternatives, individual explanation, a tested challenge process, and enforceable change controls.
- Accountable owners
- The recruiting director owns the decision. Procurement owns vendor commitments; the technical owner owns access and stop controls. Privacy and accessibility reviewers assess the revised workflow before reconsideration. A real memo must name the responsible people.
- Conditions before reconsideration
- Document the revised scope, evidence, data handling, applicant communications, responsibilities, and review process. The decision owner must explicitly authorize any next step after the required reviews.
- Recourse and stop authority
- Specify who receives applicant concerns, who can reconsider a decision, and how that person is reached. The technical owner must be able to disable the proposed use; recruiting must be able to continue through an accessible human process.
- Review triggers
- Return for review before any pilot, expansion into ranking, material vendor change, or new data use. Investigate any unexplained exclusion, inaccessible process, or inability to stop the tool before further use.
- Record status
- Illustrative redesign decision. No deployment, actual pilot, or measured outcome is reported in this example.
Use the decision-memo template for your own review, then bring the evidence and unresolved questions to the people who can make and maintain the decision.
Continue with the framework and workbook
This example applies the Authentic Intelligence Model and the distinctions described in Christopher Law’s research overview. The free companion workbook contains the Gate 0 worksheet, scorecards, and supporting governance records for a fuller review.
Get the free workbook