Productivity · Blog · 🌐 International · via agentusecases.ai
AIHR tests Claude for HR on offer letters, onboarding and performance reviews
The Claude Cowork plugin drafts HR documents, but in AIHR's pay-benchmarking test only 16% of figures were close to real market data.
- Profession
- HR Analyst, HR Professional
- Tools
- Claude Cowork, Claude for HR
- Source
- aihr.com ↗
- Country
- 🌐 International
- Found via
- agentusecases.ai (Alex Finn) ↗
- Report date
- November 5, 2026
- Verified on
- November 6, 2026
See more: AI for HR professionals
What they did
Academy to Innovate HR (AIHR) described and tested Anthropic's Human Resources plugin inside Claude Cowork. With connectors (ATS, HRIS, calendar, email, chat, policy knowledge base and compensation data) and commands such as /draft-offer, /onboarding, /performance-review, /policy-lookup, /comp-analysis and /people-report, the flow drafts offer letters (with DocuSign routing if connected), onboarding checklists, structured reviews, policy answers and people reports. AIHR itself treats the tool as a first-draft engine that still needs human review.
How it went
In its 2026 article, AIHR covers the six workflows and a pay-benchmarking test by Erik van Vulpen against the Ravio platform in the Dutch tech sector: 16% of figures were close (±5%), 23% were 5% to 15% off and 61% missed by more than 15%; at senior level underestimation reached 50%–80%. AIHR's advice is to use the plugin to draft and structure work, not as the sole source for senior pay.
Good to know
This is an AIHR test and guide, not a corporate HR case with hours saved. The article itself warns that Claude for HR senior-level benchmarking is unreliable without a dedicated compensation platform.
The prompt
I am an HR analyst at [company] and need a first draft of [offer letter / 30-60-90 onboarding plan / performance review] for [role]. Use these agreed details [paste salary, start date, goals or feedback, with no extra sensitive data] and this internal template [paste template excerpts]. Deliver the formatted text, list what still needs human review (legal, manager, compliance) and do not invent a benefit, clause or number I did not provide.
Replace what is in [brackets]. Prompt written by us from the account, not a literal copy of the source. Do not paste other people’s personal data (privacy laws such as LGPD and GDPR apply).
Summary in our own words. Results as stated in the original account; we have not tested or audited them. Read the source ↗ Case found via agentusecases.ai (Alex Finn); text and prompt rewritten by us. Found a mistake?
Similar cases
01An AI chatbot in Teams cuts time spent logging hours by 72% at dti
The Brazilian tech company built Crafty, which answers HR questions and runs internal routines for nearly a thousand employees.
02A journalist lets Claude triage his Gmail every morning
A PCWorld writer set up a routine in Claude Cowork that sorts out what matters, archives the rest, files receipts and prepares reply drafts.
03Two HR staff build an AI performance review platform without knowing how to code
At Riachuelo, two HR employees built a tool in 12 days that automates steps of the performance review cycle.
04Daily newsletters become a podcast at 5 a.m.
The agent reads the newsletters under a label, skips the weak ones, summarizes the rest and generates audio with a local voice.
05A month of sending tasks from phone to computer
A Fortune journalist used Dispatch to request email summaries and meeting prep.
06Tidying 2,200 files in the Downloads folder
A writer let Claude Cowork inventory and organize years of downloads, asking permission before deleting.
07Planning the family menu and entering recipes
The author gave Claude access to their pantry system to enter recipes and suggest the week's meals.
08Unsubscribing from newsletters in bulk in Gmail
A PCWorld journalist let Claude suggest what to cancel, asked for his approval, and then had it click the links.