I use AI every day while leading Wellhana, and the quality of the work depends heavily on what I give it to work with.
For HR, think of it like onboarding a capable employee. They need to understand the company and the job, have access to what they need, and learn how good work gets done.
I think about AI in the same three parts: context, tools and skills.
check what AI has to work with.
Context
What does it know?
Tools
What can it reach?
Skills
How should the work be done?
Context: what does it know?
Say HR wants AI to help draft benefits communication. You can open a blank chat, describe the task and ask for an email.
The result may be fine. It may also be generic, off-brand or wrong in ways that matter.
The work gets much better when AI already knows the employee population, the benefits being offered, previous communications, common employee questions, the organization’s writing style and any rules it needs to follow.
The same is true in my own work. The more AI understands Wellhana, the audience and the decisions already made, the less time I spend correcting basic assumptions.
Tools: what can it reach?
A lot of AI work still involves people carrying information back and forth manually.
Find the document, copy something into AI, get an answer, paste it somewhere else, then open another file and bring more information back.
AI becomes far more capable when it can work directly with the information and systems needed for the job: files, research, email, calendars, spreadsheets or other tools you choose to connect.
For HR, there is a big difference between asking AI to draft a benefits email from memory and letting it work from the actual plan documents, previous messages, employee FAQs and communication guidelines.
Skills: how should the work be done?
The third piece is teaching AI how your team actually works.
The first time you use AI for a task, the output may miss the mark. You review it, correct it, explain what was wrong and try again. After enough repetitions, the process becomes clearer.
Maybe benefits communication should start from the employee question, use plain language, avoid internal HR terms and end with one clear next step. Maybe a policy summary needs a fixed review process before anything goes out.
Once that way of working is clear, save the instructions, examples and standards.
That becomes a skill AI can reuse.
Does it have the right tools?
Does it know how we want the work done?
When AI gives you generic work, bad work, or gets something horribly wrong, one of those three is often missing.
Best,
Henry Kaerki
Founder, Wellhana