Boundaries

What AI should do and what you keep with humans

Boundaries are the explicit decisions about which work AI handles, which it supports, and which stays entirely with a human, no matter how capable the tool becomes.

Without deliberate boundaries, AI use tends to expand into territory where it should not be operating: consequential decisions, sensitive communications, work that requires accountability. Boundaries are not a sign of distrust in the tool. They are a sign of clear thinking about where the tool belongs.

Some boundaries are about risk: client-facing communications, financial decisions, performance reviews. Some are about quality: creative work where the standard is too high for an AI draft to be a starting point. Some are about relationships: the conversations where a human voice is the whole point. Mapping these out as a team produces an agreement everyone can hold.

Boundaries also protect the people using AI. When everyone knows what AI is and is not responsible for, individuals are not left guessing whether a decision was human or machine. That clarity matters more as AI use increases.

Make it visibleAs a team, list ten tasks you currently do and mark each one: AI handles, AI supports, or human only. The disagreements that come up during that exercise are exactly where a boundaries conversation is needed.

Why AI needs this

Each part of your work matters to AI in a specific way. Some of it is context a tool needs before it can help, some of it is work a tool can take on, and some of it is judgment that should stay with you.

Which tasks stay human

Not because AI cannot do them, but because accountability, relationships, or the stakes require a person to own the result.

Which tasks AI supports but does not decide

Many decisions benefit from AI doing the research, drafting, or analysis while a human makes the final call. Defining that support role keeps judgment where it belongs.

Which tasks AI can own

Repetitive, well-defined, low-consequence work that benefits from consistency and speed. These are the tasks where handing over fully makes sense.

Agreement across the team

Boundaries only work if the whole team holds them. AI needs the same rules applied consistently, not different standards depending on who is using it that day.

Questions to explore

Use these on your own or in a group. There are no right answers, only better conversations.

Which decisions in your work carry enough consequence that a human must make them, regardless of AI capability?
Where is AI already crossing a boundary your team has not yet talked about?
What would a stakeholder or client say if they found out a particular task was handled entirely by AI?
Which tasks feel like they should stay human but might actually be fine to hand over on closer look?
How would you tell if a boundary was being quietly ignored?

Readiness traps

Setting boundaries once and never reviewing them. As AI gets more capable and your team gets more experienced, the right boundaries shift. Review them at least yearly.
Boundaries that exist in principle but not in practice. If there is no way to enforce or check a boundary, it is not really a boundary.
Assuming that everyone on the team has the same instincts about what AI should and should not do. Those assumptions need to become an explicit, shared agreement.