11 cases
11 cases
Flag missed clock-ins and set a follow-up check for the next morning
In a hands-on test at QwenWork's public beta launch, a reporter for Chinese tech outlet APPSO had the agent spot attendance problems from clock-in and leave records, draft a group reminder and create a DingTalk to-do to check correction approvals the next morning.
HR
Source: ifanr.com · Aug 2026
Brief each interviewer, and find where candidates drop out
In the same xAI guide, one bot posts a single Slack note before each interview with who the candidate is, why the team is interested, the stage and what to focus on. Another looks across the interview loop for where candidates fail and why, and turns that into questions recruiters can ask at the first screen.
HR
Source: x.ai · Sept 2026
Source candidates overnight, then wait for approval to reach out
xAI's recruiting team describes a bot that learns from scorecards and resumes of past hired and rejected candidates, then sources new candidates overnight, adds them to the ATS as leads and drafts outreach sequences. A recruiter approves the send before the bot moves candidates to Reached Out and watches for replies.
HR
Source: x.ai · Sept 2026
Draft interview scorecards from the call transcript
xAI's recruiting team says a bot matches each interview transcript or recording to the role's scorecard and drafts the feedback form in the recruiter's voice. It then asks the recruiter a few follow-up questions, such as their rating and what to probe next, before redrafting and submitting it to the ATS.
HR
Source: x.ai · Sept 2026
Answer employee HR questions on WhatsApp and route the rest to HR
Coca-Cola Andina built an HR assistant in Microsoft Copilot Studio that answers employees' questions on WhatsApp, personalized to their role and location, and hands harder cases to an HR manager as a ticket.
HR
Source: microsoft.com · Mar 2026
Give employees one agent for HR questions and requests
Microsoft's HR team rolled out an Employee Self-Service agent inside Microsoft 365 Copilot that answers questions on leave, pay, performance and learning from vetted sources and pre-fills forms, starting country by country.
HR
Source: microsoft.com · Nov 2025
Prepare performance reviews from your existing notes in minutes
Docusign managers use NotebookLM in Google Workspace with Gemini to query their existing notes on each team member and build a performance review report, part of a company-wide Gemini rollout.
HR
Source: workspace.google.com · Mar 2026
Run a one-stop HR helpdesk agent for policy and benefits questions
Thai shopping mall developer Central Pattana built an AI agent called People Helpdesk that answers employees' HR questions and passes complex cases to the right team, as described at a Microsoft Thailand briefing.
HR
Source: news.microsoft.com · Aug 2026
Collect anonymous worker feedback by voice and summarize it for owners
US community lender Pacific Community Ventures runs AIKKA, a voice survey tool built on Claude, to gather open-ended feedback from small business workers and pass anonymized themes back to owners who have no HR team.
HR
Source: claude.com · Jul 2026
Screen frontline job applicants around the clock and route them by fit
Workforce platform Fountain built an AI Recruiter on Claude that reviews applications against hiring criteria and routes candidates, used by employers hiring hourly and frontline staff.
HR
Source: claude.com · Sept 2025
Help staff draft self-reviews from their goals and one-on-one notes
At venture firm Airtree, the Head of People wrote a detailed Claude Cowork prompt that pulls each person's goals, metrics and one-on-one notes from Notion to draft a self-review against the firm's performance framework.
HR
Source: claude.com · Mar 2026
No case matches these filters yet.
Get new cases as they land
We add cases every week. Follow the feed in any RSS reader, or check New.
How to read the evidence labels
- Vendor says
- The maker describes it on its own site or launch post. It shows what the product is meant to do, with no independent result.
- Tested by press
- A journalist or outlet ran it themselves and reported what happened, including what broke.
- User reports
- Someone using it at work posted about it. Any numbers are theirs, so we write "the poster says".
- Tested by Wize
- Our team ran it on test data and recorded the result.
Longer reads built from these cases
Our articles compare cases side by side: what worked, what broke, and where a person still decides.