The Candidate the System Ranked Last

The case

St. Mark’s is hiring a director of community ministries. The search committee receives 86 applications and worries that volunteer reviewers will favor familiar schools, church networks, and writing styles. A committee member recommends an AI screening service that can summarize résumés, compare applicants with the job description, and rank candidates. The vendor describes the process as objective and efficient.

The committee removes names before uploading materials and instructs the system to prioritize leadership, theological fit, community engagement, and communication. The top-ranked applicants have conventional ministry careers and degrees from institutions familiar to the committee. Aisha Rahman, a lay leader with extensive neighborhood organizing experience, multilingual skills, and no seminary degree, ranks near the bottom because the tool labels her experience “indirectly relevant.”

One member who knows Aisha’s work asks the committee to read her application fully. They discover that the generated summary omitted major accomplishments and described a period of caregiving as an “employment gap.” When prompted to explain the score, the tool cites qualities such as executive presence, stable progression, and polished written communication—criteria that were not in the job description. Another member points out that the church’s past hiring patterns may be embedded in both the vendor’s data and the committee’s prompts.

The chair argues that the ranking was only advisory and that humans retain the final decision. Yet the committee reviewed the top candidates first, spent more time on them, and had already begun describing lower-ranked applicants as weak. The deadline is approaching, fees have been paid, and the vendor will not reveal how scores are calculated. The committee must decide whether to restart, disclose the use to applicants, audit the process, or proceed with human review—and what faithful discernment requires when a tool has already shaped attention.

What is known—and what is not

Known so far

• The AI tool influenced the order, attention, and language of the search even though humans retain formal authority.

• Its summaries and rankings introduced criteria not stated in the job description.

• A nontraditional candidate’s relevant experience was minimized.

• The vendor’s scoring process is not transparent enough for the committee to explain or challenge confidently.

• Time, cost, and the desire to appear unbiased create pressure to continue.

Still uncertain

• How many other applications were distorted, omitted, or ranked through hidden proxy criteria.

• What data the vendor used and whether uploaded applications are retained.

• Whether applicants were informed or consented to automated screening.

• What employment, denominational, privacy, or anti-discrimination rules apply.

• Whether the committee can reconstruct a fair review without restarting completely.

Pastoral response questions

Pause before solving. Imagine that you are the leader receiving this person or community. What do you notice, what do you need to learn, and what is the next faithful step?

Locate the leadership task

1. Which parts of hiring are administrative comparison, and which are communal discernment and moral responsibility?

2. What did the committee hope the system would solve?

3. How did a ranking become influential before anyone deliberately chose to trust it?

Examine the use

1. What makes a criterion legitimate, visible, and contestable?

2. How can résumé summaries, scores, and ordering create anchoring bias even when labeled “advisory”?

3. What data should never be uploaded without clear agreements and safeguards?

Consider people, power, and trust

1. Who is advantaged by conventional language, credentials, career patterns, and access to AI-polished applications?

2. What would applicants reasonably expect to know about automated evaluation?

3. How does the church’s theological claim about vocation affect its use of opaque ranking systems?

Discern boundaries and accountability

1. Should the committee pause, restart, independently audit, or proceed differently? What values govern that decision?

2. Who has authority to challenge the tool and who is accountable for harms?

3. What must be disclosed to applicants, leaders, or the congregation?

Plan faithful practice

1. Which administrative uses might augment a search, which require caution, and which should never be delegated?

2. What review process would ensure every candidate receives a fair, human reading?

3. How should the church evaluate vendors, prompts, retention, accessibility, bias, appeals, and documentation?

Designing a tool for when you are stuck

Use these final questions to test what a practical pastoral field guide would need to provide.

1. What should a decision tool ask before AI shapes a consequential institutional choice?

2. How can it expose hidden criteria, automation bias, and the effects of ordering or summarization?

3. What must remain under accountable human and communal discernment?

Developed from the AI Pastoral Toolkit Design Sprint case corpus and Day One Synthesis and Design Charge (July 2026).