Using AI at work without getting caught out

AI tools are arriving in workplaces whether or not anybody asked for them. Used well they save real time. Used carelessly they cause the kind of mistake that is remembered for years.

The difference is a handful of habits.

Where it genuinely helps

First drafts of anything. Emails, reports, proposals, position descriptions. Getting from nothing to something rough is the slowest part of writing, and it removes that entirely. You then edit, which is far faster.

Making things shorter. Paste in three pages and ask for the key points. Reliable, because the source material is in front of it.

Changing the tone. Take a message written while annoyed and ask for a professional version. This one has saved a great many careers.

Explaining what you do not understand. A contract clause, a technical term, an acronym everyone else in the meeting seemed to know. Ask it privately rather than nodding along.

Spreadsheet formulas. Describe what you want in words and ask for the formula. Excellent at this, and it saves an enormous amount of hunting.

Preparing for something. Ask what questions you are likely to be asked, or what objections a proposal might attract. A decent rehearsal partner.

The mistakes that cause trouble

Sending output you have not read. Obvious, and it happens constantly. The writing is fluent enough that people skim it and press send. Read every word of anything going out under your name.

Trusting facts and figures. It invents statistics, dates, names and references, in exactly the same confident tone as everything else. Any number, any citation and any claim about the real world needs checking before it goes anywhere.

Pasting in confidential information. The one with genuine consequences. See what should never go into a chatbot, and check your employer’s policy before assuming.

Ignoring the policy. Many organisations now have rules about which tools are approved. Find out what yours says. “I did not know” is a poor position when the client data is already sitting on somebody else’s servers.

Letting your own skills fade. If it writes every document, your writing gets worse. Keep doing the parts of your job you want to stay good at.

The verification habit

Not everything needs checking to the same degree, and treating it all as suspect is exhausting and unnecessary. Sort it quickly:

Check hard: numbers, dates, names, legal or regulatory claims, anything you will act on, anything going to a client or your manager.

Skim: structure, tone, general reasoning. You will notice if these are off.

Do not bother: wording, phrasing, formatting. If it reads well, it reads well.

A practical trick: ask it to tell you which parts of its answer it is least confident about. It is imperfect at this, but it does tend to flag the invented bits more often than not, which is a useful prompt for where to look.

Being straight about using it

Attitudes vary and are shifting quickly, but a few things hold up.

Do not hide it. Being caught concealing it is far worse than the use itself. Nobody minds a first draft being machine-written. People mind being deceived.

Say so where the origin matters. Anything presented as your own analysis, your own recommendation, your own judgement.

Do not say so where it does not. Nobody declares spell check. A tidied-up email is not a disclosure event.

Never let it speak for you in a personal capacity. Condolences, apologies, thanks and difficult conversations should be your words. People can usually tell, and it lands badly.

The skill worth having

The people getting the most out of this are not the ones who use it most. They are the ones with a clear sense of which tasks to hand over.

That judgement comes from use. Try it on a real task, check the output properly, and notice where it was strong and where it drifted. After a few weeks you will have a reasonably accurate internal map, and you will stop either over-trusting or avoiding it.

The valuable skill is not prompting. It is knowing what to check.

If your workplace is going through this now

Team sessions work particularly well for AI, because most of the value comes out of the discussion about where it fits your specific work rather than from a demonstration.

That is something I do for small businesses around Lake Macquarie, and there is a group Introduction to AI class on Thursday evenings for individuals.