AI Guides › Judgement & Guardrails
The Difference Between A Guardrail And A Suggestion
By Nigel Guy · 2 min read
Both show up as a line of text in a skill, a prompt, or a policy document,
and they can look identical on the page. The difference is whether it's
actually enforced or just written down — and mistaking one for the other is
how a lot of "we have guardrails for that" turns out to mean nothing under
real pressure.
The rule: a guardrail is enforced regardless of pressure to skip it; a
suggestion is followed when convenient — know which one you actually have,
because they behave completely differently the one time it matters.
The mechanism for telling them apart
- Ask what happens if it's ignored. A real guardrail has a consequence
or a hard stop built in — a required confirmation, a blocked action, a
review step that can't be skipped. A suggestion has no such mechanism; it
relies entirely on someone choosing to follow it.
- Check whether it survives urgency. The real test of a guardrail is
whether it holds up when someone's in a hurry and tempted to skip it.
A rule that quietly gets bypassed "just this once, we're behind" was
never actually a guardrail.
- Notice vague, unenforceable language. "Use good judgement," "be
careful with sensitive data" — these read like guardrails but function as
suggestions, because nothing about them is checkable or enforced.
- Build real guardrails as mechanisms, not reminders. A required
confirmation step, a permission that has to be explicitly granted, a
hard limit on scope — these hold under pressure in a way that written
advice doesn't.
What to skip
Skip writing a guardrails section full of things that sound serious but
have no actual enforcement behind them — it creates false confidence that
something is protected when it isn't. And skip assuming a documented
suggestion will be followed consistently just because it's written down and
everyone agreed with it in the moment.
Guardrails
- The gap between a guardrail and a suggestion is most dangerous exactly
where the stakes are highest — audit your real guardrails there first, not
your low-stakes processes.
- A guardrail that's inconvenient enough gets quietly worked around unless
it's genuinely built into the mechanism, not just the documentation.
- This applies to your own personal rules as much as to skills or team
processes — a rule you've written down but never actually enforce on
yourself is a suggestion, whatever you call it.
All 751 AI guides · JulieMango plans from £17/mo