The Operations Bridge

Point seven in the AI Strategy Points series. Last time was about starting with data instead of training. This post is about a metaphor that has done more for how I manage AI’s failure modes than any technical explanation ever has.

The metaphor

AI is a super intelligent, hyper-literal, eager, but inexperienced intern who will produce exactly what you ask for — even if it’s wrong. Every word in that sentence is doing work, and I’ve found it more useful for setting expectations than any technical explanation of how a language model actually functions.

Super intelligent: it has access to more information than almost anyone you’ve ever managed, and it can synthesize that information fast. This is genuinely impressive and genuinely useful.

Hyper-literal: it will do precisely what you asked, not what you meant. If your instructions had a gap or an ambiguity, it will fill that gap with something plausible rather than come find you to ask a clarifying question, the way a cautious human intern might.

Eager: it will always produce an answer. It doesn’t get tired, doesn’t push back out of frustration, and rarely says “I’m not confident about this” unless you explicitly ask it to assess its own confidence.

Inexperienced: it has no scar tissue from your specific operation. It hasn’t lived through the changeover that went wrong last March or the workaround your team quietly developed because the official process didn’t fit reality. It knows patterns, not your plant.

What this means for how you manage it

You wouldn’t hand a brand-new intern a vague instruction and walk away assuming they’d fill in the gaps correctly. You’d give clear direction, check their early work closely, and gradually extend trust as they demonstrated they understood the context. That’s exactly the posture I now bring to working with AI.

The mistake I see most often is people treating AI outputs the way they’d treat advice from a trusted senior colleague — assuming context and judgment that simply isn’t there. The hyper-literal, inexperienced parts of the metaphor are the ones people forget under time pressure, and that’s exactly when a confidently wrong answer slips through unchallenged.

The eagerness is the part that catches people off guard

A human intern who doesn’t understand an assignment will often hesitate, ask a question, or produce something visibly unfinished. AI rarely signals uncertainty that way. It will hand you a fluent, complete-looking answer whether the underlying reasoning was solid or shaky. That eagerness is genuinely useful for speed and genuinely dangerous if you mistake fluency for accuracy.

A question for leaders

Think about how you’d manage a talented, eager, but brand-new intern on a task that actually mattered. Are you applying that same level of oversight to what AI produces, or are you extending trust it hasn’t earned yet?

Next: scaling AI is about successful test cases — and what that has in common with Kotter’s approach to change.