By Johannes Sundlo — People AI Evangelist. I help leaders and HR teams turn AI into adoption that sticks, through keynotes, workshops and change programs.
Most HR teams don’t fail at AI because the tools are too hard. They stall because AI gets bolted onto the side of the work instead of built into it. This guide is the practical version — where to start, the exact steps to roll it out, the use cases worth your time, and how to know it’s actually working. No hype, just adoption that sticks.
Where HR teams actually start with AI
Don’t start with a strategy deck. Start with the tasks your team already repeats every week — the ones that are text-heavy, rule-based, and low-risk. That’s where AI gives value on day one and builds the confidence you’ll need for the harder stuff later.
- Drafting: job ads, interview guides, policy first drafts, onboarding emails, performance-review summaries.
- Summarising: long policy documents, survey free-text, exit-interview notes, meeting transcripts.
- Structuring: turning messy notes into competency frameworks, comparing candidates against criteria, tidying spreadsheets.
- Answering: a private assistant for “how does our parental-leave policy work?” type questions, grounded in your own documents.
A 5-step playbook to roll out AI in HR
- Pick one painful, repeated task. One. Make it something a real person does every week and quietly dislikes. Specific beats ambitious.
- Write the prompt as a process, not a question. Give the AI context (your tone, your policy, an example of good output) and constraints. A reusable prompt is an asset — save it.
- Put a human in the loop. AI drafts, a person decides. Make that rule explicit so people trust the output and own the result.
- Make it the default, not an extra. Adoption happens when the AI step lives inside the existing workflow — the template, the checklist, the shared doc — not in a separate tool people forget to open.
- Capture the win and tell the story. “This used to take 40 minutes, now it takes 8.” Concrete, local proof is what spreads AI across a team — far more than a mandate from above.
High-value AI use cases in HR
- Recruiting: first-draft job ads in your voice, structured candidate comparisons against your criteria, faster screening of free-text applications.
- Onboarding: personalised onboarding plans, an always-on assistant that answers new-hire questions from your handbook.
- People analytics: turning engagement-survey free-text into themes in minutes instead of days; spotting patterns you’d otherwise miss.
- L&D: generating role-specific learning paths, drafting workshop content, summarising what “good” looks like for a competency.
- Capturing hidden knowledge: structured interviews with senior staff, turned into reusable documentation before that expertise walks out the door.
The mistakes that kill HR AI projects
- Boiling the ocean. A 12-month “AI transformation” with no win in the first month loses the room. Ship something small and real first.
- No data guardrails. Decide early what can and can’t go into which tool. Sensitive employee data needs a clear, simple rule everyone understands.
- Tool-first thinking. Buying a platform before you’ve nailed one workflow. The workflow is the product; the tool is just plumbing.
- Treating it as IT’s job. AI in HR is a change-management problem wearing a technology costume. People adopt habits, not software.
How to know it’s actually working
Skip vanity metrics. Track three things: time saved on the target task, how many people use it without being told to, and quality (does the human editing get lighter over time?). When usage spreads on its own and edits shrink, you’ve got real adoption — not a pilot that quietly died.
The fastest way to make this happen in your team
Reading a guide is one thing; getting a whole HR team to change how they work is another. That’s what I do — keynotes that make AI click for the room, and hands-on workshops where your team leaves with working prompts and habits they’ll actually use the same week.
Bring practical AI to your HR team
Keynotes and workshops that turn AI curiosity into adoption that sticks. Let’s find the format that fits your team.
Want to talk it through first? Get in touch and tell me where your team is today.
Common questions about AI in HR
How do I start using AI in HR?
Start with one repeated, low-risk, text-heavy task your team already does every week — drafting job ads, summarising survey free-text, or answering policy questions. Get one real win, save the prompt, then expand. Don’t begin with a platform or a strategy deck.
Is it safe to use AI with employee data?
It can be, with a clear rule. Decide up front what data may go into which tool, and keep anything sensitive out of public AI tools unless you have an enterprise agreement that guarantees your data isn’t used for training. One simple, well-understood guardrail beats a long policy nobody reads.
Which HR tasks should I automate first?
The high-frequency, low-risk ones: first-draft job ads, interview guides, onboarding emails, policy summaries, and turning engagement-survey comments into themes. Keep a human in the loop — AI drafts, a person decides.
How do I get my HR team to actually adopt AI?
Build the AI step into the existing workflow instead of adding a separate tool, put a human in the loop so people trust it, and make local wins visible (“this used to take 40 minutes, now 8”). Adoption spreads through proof, not mandates. A hands-on workshop accelerates this dramatically.
Related guide: Leading the change from the top? Read AI for Leaders: a practical guide to driving adoption.
Keep reading: AI adoption that sticks · AI in recruiting · ChatGPT prompts for HR · running an AI pilot.
Where should HR start with AI?
| Task type | Examples | Why start here |
|---|---|---|
| Drafting | Job ads, policy drafts, onboarding emails | High frequency, low risk |
| Summarising | Survey free-text, long policies, notes | Big time savings, text-heavy |
| Answering | Policy Q&A grounded in your own docs | Daily and repetitive |
| Structuring | Notes → scorecards, competency frameworks | Clear input, human decides |
