USING AI? DECIDE WHAT STAYS HUMAN
Lu Pillar: Design Sustainable Work
Read time: 3 to 4 minutes
IF YOU ONLY DO ONE THING
Pick one real task where AI is being used and answer four questions:
What can AI help with?
What still needs a person to check or decide?
What rules already apply?
What do you need someone else to clarify?
You don’t need to solve AI for the whole organisation. Start with the work happening in front of you.
WHAT'S HAPPENING
AI use is often moving faster than workplace guidance.
Some people are already using it. Others are avoiding it. Different people may be using different tools, in different ways, with different ideas about what is acceptable.
If expectations are unclear, people fill in the gaps themselves.
That makes it harder to see where AI is genuinely helping, where it is creating risk and who is responsible for the final work.
WHAT IT CAN LOOK LIKE
Different people use AI differently for the same task.
Nobody is sure what the organisation allows.
AI-generated work is treated as finished without a clear quality check.
“AI should make that quick” becomes the expectation without understanding what still needs human work.
People keep their AI use quiet because they are unsure how it will be received.
Work that used to help people learn disappears without another way to build that capability.
WHAT IT COSTS
When AI use is unclear:
Polished work can still be wrong.
You may not know how work was produced or checked.
Important learning and development opportunities can disappear without a replacement.
People may use tools in ways the organisation has not approved, including placing confidential information into applications without security.
Useful applications can also stay hidden because nobody is talking about what works.
Unclear AI use can create both avoidable risk and resistance.
WHAT THIS ENABLES
Use AI where it improves the work, while keeping people responsible for understanding, checking and owning the outcome.
You don’t need to be an AI expert to manage this well. But you do need to make clear what you know, what the team is responsible for and what needs to be clarified by someone with the right authority or expertise.
When expectations are clear:
People are more open about how they use AI.
Work is checked more consistently.
Responsibility for the final output stays visible.
Managers can see where AI is actually saving time.
Risks and gaps are easier to identify early.
Learning and judgement are less likely to disappear by accident.
The team can improve its use together instead of everyone developing their own approach.
Good AI use is not about using more of it. It is about knowing where it helps and where a person still needs to think, check and decide.
USE THESE WORDS
When talking with your team, try:
“Before we keep using AI for this task, let’s make the expectations clear. We’re using it to: [draft / summarise / generate options / analyse / other]. You still need to: [check facts / apply judgement / make the decision / review the final work].
The rules we already know are: [approved tools / information restrictions / existing policy].
What we still need clarified is: [question]. I’ll take that to [IT / privacy / legal / manager / relevant owner].”
Then write the agreed approach down where the team can easily see and find it.
If you don’t know what the organisation allows, don’t guess. Try saying:
“I don’t have a clear answer on what is approved here, so I’m going to get one before we make this normal practice.”
Then take the specific question to the person who owns it as soon as possible.
TRY IT THIS WEEK
Choose one task where AI is already being used or seriously considered.
Agree what AI is helping with, what a person still owns and what questions need to be escalated.
Start with one task, not the whole organisation.
CHECK IN TWO WEEKS
Ask the people doing that work:
What are we using AI for? What still needs a person to check or decide?
If you get different answers, the expectation is still unclear.
WHEN THIS GUIDE ISN'T ENOUGH
Privacy, confidentiality, intellectual property, contracts, cyber security, industry requirements and legal obligations require appropriate specialist advice and organisational controls. Lu focuses on the people and work design side of technology: purpose, roles, judgement, learning, decision-making and accountability. We don’t select technology, build automation or provide legal, cyber, privacy or technical governance assurance.
NEED MORE SUPPORT?
Lu offers two next-step options.
Run It gives teams practical tools to work through the issue themselves.
Embed It provides independent support to understand what is driving the problem, tailor the response and help new practices stick.
