The Honest Guide to Using AI at Work, by Industry

The Honest Guide to Using AI at Work, by Industry

Most "AI at work" advice is written as if everyone has the same job. They don't. A teacher, a paralegal, a nurse, and a marketing manager can all use the same model — and should use it in almost completely different ways, with completely different lines they shouldn't cross.

This is a short, honest tour of how that plays out in a few fields. For the full set — education, legal, healthcare, marketing, software, journalism, students, customer support, and creative work — see the industry playbooks, where each field gets its own deep dive.

The one thing that's true everywhere

Before the differences: the single universal rule is that AI is a drafting and thinking partner, not an authority. Every field's version of "don't trust it blindly" looks different — a lawyer checks citations, a developer runs the code, a teacher checks the reading level — but the shape is the same. The model is fast, tireless, and confident. It is not careful. Careful is your job.

Education: a planning superpower with an integrity problem

For teachers, AI is genuinely transformative for the invisible work: differentiating materials across reading levels, generating practice questions, drafting parent communications, building rubrics. An hour of Sunday planning can compress to twenty minutes.

The tension is obvious: the same tool that helps a teacher plan is the tool their students use to skip the assignment. The schools handling this well aren't banning or ignoring it — they're redesigning assessments (more in-class work, more oral defense, more process-graded assignments) and teaching students to use AI openly rather than pretending it doesn't exist.

Legal: real leverage, real malpractice risk

Lawyers get enormous value from AI in research summarization, first-draft contract review, and translating legalese for clients. Document review that took a junior associate a week can take an afternoon.

But legal is the field with the sharpest documented failure mode: models invent case citations that look completely real, and lawyers have been sanctioned for filing briefs containing them. The professional rule that emerged is simple and absolute — nothing cited by a model enters a filing until a human has pulled the actual source. Client confidentiality adds a second line: what you paste into a consumer chatbot may not stay private, so firm-approved tools and redaction matter.

Healthcare: administrative relief, clinical caution

The clearest win in healthcare is the paperwork. Clinicians spend a staggering share of their day on documentation, and AI scribes and drafting tools genuinely give hours back — time that goes to patients instead of screens.

Clinically, the posture is different. Models can be useful for differential-diagnosis brainstorming and patient-friendly explanations, but they have no access to the chart, no liability, and no license. The line most health systems draw: AI may draft and summarize, but a clinician decides, and anything patient-facing gets read by a human before it's sent.

Marketing: volume is solved, distinctiveness isn't

Marketing teams were the fastest adopters, and the results are mixed in an instructive way. AI is excellent at the volume work — variants, repurposing, SEO drafts, social captions — and mediocre at the thing that actually differentiates a brand: a point of view.

The teams getting real value use AI for the first 80% and humans for the last 20% — the angle, the voice, the specific customer insight no model has. The teams that automated everything discovered their content now reads exactly like every competitor who did the same. (If that sounds familiar, our piece on why AI drafts sound generic covers the tells, and the Voice Writer is built to fix exactly that problem.)

The pattern across all of them

Look across the fields and a consistent shape emerges:

• AI is strongest on the work around the work — drafts, summaries, variants, first passes — and weakest at the moment of actual judgment.
• Every field has a hard line where "AI-assisted" becomes "professionally irresponsible," and the line is different everywhere. Knowing yours is part of using the tool well.
• Disclosure norms differ. What's fine in marketing copy is a firing offense in journalism and an academic-integrity case in a classroom. When in doubt, ask what the expectation is before you're caught guessing it. Our ethics hub digs into the underlying questions — bias, attribution, transparency — that sit underneath all of these.

Get the full playbook for your field

These are four-minute sketches; the real playbooks go much deeper — specific workflows, the tools people actually use, the mistakes to avoid, and ready-made agent templates for roles in each industry. Find yours at the industry hub: education, legal, healthcare, marketing, software development, journalism, students, customer support, and creative work.