AI in HR in 2026: what's actually working, with examples
Daniele D'Esposito, Founder, Org360
, 9 min read
In plain English
AI in HR is real but uneven. SHRM's 2026 research found 39% of organizations use AI in their HR function. Where it works, it does narrow, repetitive jobs: answering policy questions, drafting, summarizing survey comments, preparing reviews. IBM's AskHR handles common employee questions end to end 94% of the time.
The new development in 2026 is connectors: HR platforms such as Lattice, Culture Amp, Leapsome and 15Five now let Claude or ChatGPT reach your people data. The limits are clear too: people decisions stay with people, and Europe's AI Act treats AI used for hiring and managing workers as high-risk.
Every HR conference this year has been about AI. Strip out the announcements and a simpler picture appears: AI is good at a few specific HR jobs, and those jobs are spreading fast. Here's what the data and the real examples show.
How much HR actually uses AI
- 39% of organizations use AI in their HR function, and another 23% use it elsewhere in the business, according to SHRM's State of AI in HR 2026, based on 1,908 HR professionals.
- 41% of workers use AI in their work, and just under half of them (44%) call their own output "AI slop", in SHRM's survey of over 5,000 workers published in June 2026.
- 45% of managers say AI has improved their team's work as much as they expected, Gartner found. Managers experiment more than staff (46% against 26%), and only 7% of organizations give guidance on how to use the time AI saves.
Read together: adoption is real but far from universal, quality is uneven, and the time AI frees up mostly disappears unless someone decides what it's for.
Where AI is working in HR
Answering employees' questions
IBM's AskHR is the best documented example. It automates more than 80 HR tasks, from payslips and leave requests to managers' transfers and organization changes, and handled more than 11.5 million employee interactions in 2024. IBM reports a 94% containment rate for common questions, meaning no handover to a person. Human advisors handle the complex cases.
Reading what employees say
Engagement platforms now summarize thousands of survey comments in seconds. Microsoft's Copilot in Viva Glint summarizes up to 8,000 comments on a chosen theme; Culture Amp shows the comments behind each AI summary. We cover this in how engagement survey tools use AI.
Preparing reviews and feedback
Lattice, 15Five, Culture Amp and Workleap all use AI to help managers draft reviews from a year of goals, notes and feedback. Assessment firms are moving too: Hogan bought Convogo in July 2026 to give certified practitioners an AI first draft of an interpretation, and CCL uses AI trained on expert coders to give leaders behavioral feedback. Details in how 360 tools use AI.
Connecting HR data to the AI people already use
The quiet shift of 2026 is the HR connector. By November 2026, Lattice, Culture Amp, Leapsome, 15Five, Workleap, Betterworks and BambooHR (both in beta) all published one, and so does Org360. A manager asks Claude or ChatGPT, and the answer draws on the HR platform within that manager's own permissions. Most connectors read freely and write only where the vendor lists a specific action.
Where the limits are
- Decisions about people. Every serious vendor we read, from Culture Amp to Hogan to EchoSpan, says the AI supports a human decision and doesn't make it. EchoSpan's InsightScout won't advise on hiring, firing or pay.
- Regulation. The EU AI Act classes AI used in employment and worker management as high-risk. The Digital Omnibus, approved in June 2026, moves those obligations to 2 December 2027, while transparency duties under Article 50 applied from 2 August 2026. In New York City, Local Law 144 has required bias audits of automated hiring and promotion tools since July 2023.
- Quality. SHRM's "AI slop" figure is a reminder that a draft is a draft. The teams getting value check the output before it reaches an employee.
What this means for a small HR team
- Start with reading, not deciding: summaries of survey comments, policy answers, first drafts.
- Before connecting your AI to people data, check what it can read, what it can change and whether confidentiality rules still apply.
- Write one rule down: AI output is never the only basis for a decision about pay, promotion or discipline.
- Decide what the time saved is for, or it will vanish.
Where Org360 fits
We make Org360, a survey tool, so this is one small corner of the picture. Org360 uses AI to write the report from 360 and engagement surveys, with names removed first, and lets you run surveys from Claude, ChatGPT or Gemini through its connector, including setting up and sending one after you confirm. It makes no decisions about people. See how the connector works.
Frequently asked questions
How many companies use AI in HR?
SHRM's State of AI in HR 2026 found 39% of organizations use AI in their HR function, with another 23% using it elsewhere in the business.
What are examples of AI in HR?
Answering employee questions (IBM's AskHR resolves 94% of common questions without a person), summarizing survey comments, drafting performance reviews, and connectors that let Claude or ChatGPT read HR data within a user's permissions.
Is AI in HR regulated?
Yes, increasingly. The EU AI Act treats AI used for employment and worker management as high-risk, with obligations moving to 2 December 2027 under the Digital Omnibus. New York City requires bias audits for automated hiring and promotion tools.
What is an HR MCP server?
A connector that lets an AI assistant such as Claude or ChatGPT read data in your HR software, and sometimes make listed changes, using the signed-in person's permissions.
Sources
All checked on 15 November 2026. Vendors change these pages often, so check the current version before you decide.
- SHRM: AI in HR 2026, from hype to measured impact
- SHRM: The State of AI in HR 2026
- SHRM: Navigating AI in the Workplace 2026
- Gartner: 45% of managers say AI met expectations
- IBM: AskHR case study
- Lattice MCP integrations overview
- Culture Amp MCP
- Leapsome MCP
- 15Five MCP feature overview
- Workleap: the agent in Claude and ChatGPT
- Betterworks MCP
- BambooHR MCP server (beta)
- Microsoft: Copilot in Viva Glint
- Hogan: acquisition of Convogo
- EchoSpan: AI insights toolkit
- Jones Walker: EU AI Act high-risk delay and Article 50
- NYC DCWP: Automated Employment Decision Tools
Get new research by email
About one email a month, when we publish something new on coaching, 360 feedback and engagement surveys.
We'll email you once to confirm. Unsubscribe anytime. See our data protection policy.