An AI HR assistant is an AI tool that does routine HR work on request: it answers employee questions from your policies, drafts documents, summarises feedback and prepares the paperwork around hiring, onboarding and reviews. A person still checks the output and makes every decision about people.
There are three common kinds. A general assistant such as ChatGPT, Claude or Copilot, set up with your company's documents. An assistant built into HR software, such as Ask BambooHR. And a bot that lives in Slack or Microsoft Teams and handles one job, like collecting feedback or answering leave questions.
This guide covers what an AI HR assistant can do, how it differs from a chatbot or an agent, which kind fits a small HR team, and how to set one up without putting employee data at risk.
AI HR assistant vs HR chatbot vs AI agent
The terms overlap, and vendors use them loosely. A practical way to tell them apart:
- HR chatbot: answers questions. It works from a fixed set of answers or from your handbook. It doesn't do anything beyond replying.
- AI HR assistant: answers questions and produces work. It drafts a policy, summarises ten peer reviews, builds an onboarding plan. You ask, it delivers, you check.
- AI HR agent: carries out steps on its own across tools. It might send reminders, update a record and create a task. More useful, and more to check, because it acts rather than drafts.
"AI HR teammate" is a newer label for the same idea: an assistant that sits where the team already works, usually Slack or Teams, and handles recurring HR tasks without someone opening a separate tool.
For most HR teams at 30 to 1,000 people, the assistant is the right starting point. It saves time on writing and repetitive questions while keeping a person between the AI and the employee.
What an AI HR assistant can do
Answer employee questions
Leave balances, benefits, expense rules, who to ask about what. These questions arrive daily and the answers already exist in your handbook. An assistant that reads the handbook can answer them and point to the source, and HR stops being the search engine for the company.
Ask BambooHR is one example: BambooHR says it answers employee questions about benefits, policies and more from data and documents stored in BambooHR, and respects existing role-based permissions so people only see what they're entitled to.
Draft HR documents
Job descriptions, offer letter templates, policies, announcements, interview kits. The assistant writes the first draft from your notes; you edit. For ready-to-use prompts by task, see ChatGPT prompts for HR.
Onboarding
A task list for each new hire with owners and dates, a welcome email, a first-week schedule and a 30/60/90-day plan for the manager to edit. The assistant needs your onboarding process written down once; after that it can produce a plan per hire in minutes. Effy AI's free skill for an onboarding plan for a new hire shows the setup.
Performance reviews and feedback
Review forms for a role, written summaries of self, manager and peer feedback, rewrites of vague comments into specific ones, and checks for biased wording before a review is shared. This is where an assistant saves managers the most time, and where it needs the closest checking. It should never set the rating.
Dedicated tools do this without prompting. They're compared in performance appraisal software, and if your team lives in Slack, in performance reviews in Slack.
Development and career plans
Individual development plans from a review's development areas, skills matrices, and drafts of a career framework from your job descriptions. The platforms that run these are compared in employee development software.
Surveys and people data
Grouping hundreds of open survey comments into themes, drafting manager action plans from results, and answering plain-language questions about headcount or attrition when it has access to the data.
What an AI HR assistant shouldn't do
Let it prepare; let people decide. Keep these with a person:
- Hiring, firing and promotion decisions
- Ratings and pay decisions
- Disciplinary steps and performance improvement plans
- Conversations about health, family, conflict or someone's performance problems
- Final answers on employment law
An assistant also states wrong things confidently. Anything it says about law, pay or a named employee needs checking before it goes anywhere.
Which kind of AI HR assistant fits your team
A general assistant (ChatGPT, Claude, Copilot) fits if most of your HR work is writing, summarising and answering questions, and you're happy to set it up yourself. It's the cheapest way to start and the most flexible. The work is in giving it context: your policies, your processes, your tone.
An assistant built into your HRIS fits if employees mostly ask about data that lives there: leave, benefits, org chart. It already has the data and the permissions. It won't help much with drafting reviews or onboarding plans unless the vendor built that in.
A specialist tool with AI built in fits a specific process you run repeatedly, such as performance reviews, 360 feedback or engagement surveys. You get the AI without writing prompts, inside a workflow that already handles reminders, deadlines and reports. The wider market is compared in AI HR software.
Many teams end up with a mix: a general assistant for writing, their HRIS for employee questions, and a specialist tool for reviews.
How to set up an AI HR assistant
- Write an AI use policy first. Which tools are approved, what data never goes in, and when output needs a person to review it.
- Pick the plan with the right data terms. On ChatGPT's individual plans (Free, Plus and Pro), OpenAI says it may use your conversations to train its models unless you turn that off in Settings > Data controls. On ChatGPT Business and Enterprise, it does not train on your data by default. Check the equivalent terms for any tool you use.
- Give it context once. A short set of documents: what the company does, how many people, where they work, your handbook, your review process, how you write. An assistant that reads this before every task stops producing generic answers. Effy AI's free skill for HR context files builds this set for you: company basics, one file per HR process, and no employee names.
- Start with one repeated task. Onboarding plans or review summaries are good first choices: frequent, time-consuming and easy to check.
- Write down how the task is done. A page on the steps, the inputs and what good output looks like. Save it so the assistant follows it every time instead of you re-explaining.
- Run it for a month and compare. Time spent and quality against the old way. Then add the next task.
The free AI for HR course follows these steps with Claude, from setting up context to saving reusable skills.
Risks and rules
The same four risks come up with any AI used in HR: bias carried over from past decisions, privacy of employee data, confident errors, and employees not knowing when AI is involved.
Rules are tightening around AI that influences decisions about people. The EU AI Act treats AI used in recruitment and in decisions about promotion, termination and performance evaluation as high-risk, and New York City and Illinois already regulate AI in employment decisions. An assistant that drafts and summarises, with a person deciding, is the lower-risk use. If you use AI in screening, ratings or promotions, check with employment counsel first. This is a summary, not legal advice. The rules HR teams run into most are covered in AI for HR.
Where Effy AI fits
Effy AI, our own product, is a performance management tool, not a general HR assistant. It uses AI for one job: performance reviews. It generates review forms and written summaries, runs bias checks and quality nudges on written feedback, and works inside Slack. It's free for up to 5 users and starts at $3 per seat a month. Under 300 employees, the options are compared in performance management software for small business.
The short version
An AI HR assistant answers routine questions and drafts HR work so your team spends less time on admin. Start with a general assistant and one repeated task, give it your company's context once, keep employee data out of tools without the right terms, and keep every decision about a person with a person.
