AI strategy: HR specialist

Prioritized AI adoption map. Date: 2026-06. Sources: GDPval, AEI, O*NET, and profession-automation-2026.md. ← Role hub

Context
AI is common in HR, and screening can save substantial time. AEI and O*NET adoption estimates are only a guide. Hiring decisions and culture stay with people. Bias is a serious risk.

1. Recruiting (high priority)

No.	Goals	KPIs	Initiatives
1	Screen and rank resumes	time to hire · shortlist quality	ATS with AI · confidence bands, which show how certain the system is · review borderline cases manually
2	Source candidates	coverage · relevance	semantic search with Potok, Huntflow, or hh.ru

2. Coordination (high priority)

No.	Goals	KPIs	Initiatives
3	Schedule interviews and communicate	drop-off · speed	scheduling bots · automatic candidate updates

3. Content (medium priority)

No.	Goals	KPIs	Initiatives
4	Job posts, letters, and policies	time · quality	drafts with an assistant · human removes cliches
5	Onboarding and a policy-based knowledge base	onboarding time	NotebookLM based on policies · AI answers for new hires

4. Ethics and decisions (foundation)

No.	Goals	KPIs	Initiatives
6	Control bias and protect personal data	incidents · audits	ban automatic rejection · audit algorithms · approved data environment

Where to start: the first round

The tables are a map of what is possible, not your plan. Start with process scoring. Ask whether the task repeats, has enough volume, produces a checkable result, carries an affordable error cost, and has usable data. For most HR teams, the first round looks like this:

No. 3, interview scheduling and communication. It repeats for every candidate and has enough volume. Compare the output with the calendar. An error costs a rescheduled slot, not a hiring decision.
No. 2, candidate sourcing. A recruiter can see whether the results are relevant, and a person makes the final decision to put someone into the pipeline.
No. 4, job posts, letters, and policies. A person always edits the draft, checks the terms, and removes cliches before publication.

Not in the first round: No. 1, "resume screening and ranking," has the highest error cost even though the table labels it high priority. Biased automatic rejection is a red flag in the role hub. Do not start without confidence bands and regular algorithm audits. No. 5, "onboarding and the knowledge base," works only when the policies are already in order. Otherwise, AI will answer from bad material. No. 6, bias control, is a continuing check across all initiatives, not a separate task for later.

Keep these parts of HR human

The hiring decision. It is final and cannot be delegated at any maturity level.
Rejection feedback. A person chooses the words and tone, with the legal risks of an unjustified rejection in mind.
Culture, conflict, and difficult conversations. These shape trust inside the company.
Assessing personality and soft skills in an interview. This requires judgment from a live conversation, not just text.
Candidate and employee personal data. Keep it inside an approved environment and own the consequences of a leak.

Review points and stop thresholds

Initiative	What a person checks	Stop threshold (example, replace with your own)
No. 3 interview scheduling	compare slots and reminders with the real calendar and candidate replies	if a candidate is lost because communication failed, return to manual message control
No. 2 sourcing	check relevance and look for skew against any candidate group	a systematic drop for one group means auditing the search criteria
No. 4 job posts and letters	remove cliches and verify terms before publication	an inaccurate term or cliche in a published role means pausing publication and manually reviewing the next batch

Before you trust the flow, build a reference set of 20 cases: 20 resumes with a known human hiring decision. Test every new screening tool against them first.

Two paths from here

You are an HR specialist: open the role hub, then use the Volume 1 workbook to move from your week to an audit and a 90-day plan.
You lead the HR function: use the Human Review Matrix, write the first agent contract with boundaries for automatic rejection, personal data, and escalation, then continue with the Volume 2 workbook.

Discipline
Do not let AI reject candidates automatically because bias can hide in the system. A person owns hiring decisions, feedback, and culture.

→ Role hub · audit · plan.

Sources: profession-automation-2026.md; book, Chapters 2, 3, and 7. Version: 2026-07-09. What changed: added "Where to start: the first round," "Keep these parts human," "Review points and stop thresholds," and "Two paths from here."