Process Agent — AI-Assisted Job Description Optimization Workflow
What This File Covers
This file defines where and how AI tools can be used in producing position profiles, job descriptions, and job ads. Every AI-generated output requires practitioner review before it becomes a deliverable.
Design Phase — Building the Deliverables
Drafting Position Profiles from Extraction Notes
When to use AI: After the extraction interview is complete and notes are organized. AI drafts the position profile from the extraction content.
What to provide:
- Organized extraction notes
- Must-have and nice-to-have requirements as designated during extraction
- Organizational context from the interview
- Compensation range
- Reference data for proper nouns
Expected output: A draft position profile following the structure in 03a-golden-example-consultant.md.
Quality check:
- Does the role purpose explain WHY, not just WHAT?
- Are must-haves genuinely non-negotiable, or did AI soften the language?
- Is the organizational context specific and current, not generic?
- Are responsibilities written as actions, not competencies?
- Does every element trace to something said in the extraction?
Common AI failure modes:
- Inflating nice-to-haves into must-haves (AI tends to treat everything as required)
- Generic organizational context ("a dynamic, growing organization")
- Competency-language responsibilities ("provides strategic leadership" instead of actionable descriptions)
- Adding requirements not in the extraction (AI fills gaps with industry-standard language)
- Overly formal tone that doesn't match how the stakeholders described the role
Drafting Job Descriptions from Position Profiles
When to use AI: After the position profile is finalized. AI drafts the formal job description.
What to provide:
- The finalized position profile
- Any compliance requirements (EEO statement, salary disclosure requirements)
- Format specifications
Expected output: A draft job description.
Quality check: Cross-reference against the position profile. Same must-haves? Same responsibilities? Nothing added, nothing dropped?
Drafting Job Ads from Position Profiles
When to use AI: After the position profile is finalized and the target candidate is defined. AI drafts the attraction-oriented job ad.
What to provide:
- The finalized position profile
- The target candidate profile (who are we trying to reach, what motivates them)
- The publication channels (the tone may vary by channel)
- The organization's compelling elements from extraction
- Tone guidance
Expected output: A draft job ad that leads with compelling elements and reframes requirements as identity.
Quality check:
- Does it open with what makes the opportunity compelling, not a requirements list?
- Is it written from the candidate's perspective?
- Does it accurately represent the role (no promises the position profile doesn't support)?
- Would the target candidate recognize themselves in the "Who You Are" section?
- Is the tone right for the channel and audience?
Common AI failure modes:
- Writing an ad that reads like a reformatted job description
- Using generic attraction language ("exciting opportunity in a fast-paced environment")
- Over-promising or adding compelling elements not grounded in the extraction
- Corporate tone when the organization's culture is informal (or vice versa)
- Requirements section that reverts to list format instead of identity framing
Requirement Inflation Analysis
When to use AI: When a prior job description exists and needs to be evaluated before extraction.
What to provide:
- The prior job description
- The role title and level
Expected output: A flag report identifying likely inflated requirements — qualifications that seem disproportionate, accumulated certifications, excessive experience thresholds, requirements that may have been added by successive hiring managers without validation.
Quality check: The practitioner uses this analysis to inform the extraction conversation, not as a final determination. Only the stakeholders can confirm what's truly required.
Cross-Deliverable Consistency Check
When to use AI: After all deliverables are drafted.
What to provide: All three deliverables.
Expected output: A comparison identifying any inconsistencies — must-haves that appear in one document but not another, role purpose language that diverges, compensation information that differs.
Quality check: The practitioner reviews every flagged inconsistency and resolves it by updating the deliverable that's wrong. When in doubt, the position profile is the source of truth.
Where AI Cannot Replace the Practitioner
Must-have designation. The distinction between must-have and nice-to-have requires organizational judgment informed by stakeholder conversations. AI can suggest; only the practitioner confirms.
Role validation. Whether the role should exist is a strategic conversation. AI cannot navigate organizational politics or assess strategic fit.
Stakeholder disagreement resolution. When the hiring manager and their leader disagree about the role, the practitioner facilitates resolution. AI cannot mediate.
Organizational context. The "what's happening right now" section requires insider knowledge. AI can draft from extraction notes, but the practitioner must verify specificity and accuracy.
Tone calibration. The job ad's voice must match the organization and the target audience. AI approximates; the practitioner decides.
Compensation decisions. AI cannot determine what the organization should pay. Compensation benchmarking is a separate analytical process.
What AI Does Not Do
Invent requirements. If the extraction didn't capture a requirement, AI does not add one from industry knowledge. Missing requirements are gaps, not AI fill opportunities.
Replace the extraction interview. AI can draft from notes, but it cannot conduct the stakeholder conversation that produces those notes. The extraction is human work.
Guarantee legal compliance. AI can flag common issues (protected-class language, missing EEO statements) but cannot provide legal advice. Compliance review is the client's responsibility, informed by the practitioner's recommendations.
Make the ad compelling through language alone. If the role and organization aren't genuinely compelling, clever writing won't fix it. The compelling elements must be real — AI can present them well, but it can't create them.