Good answers from AI are easier to get than you think, and so is finding new ways to put it to work.
AI is well into our lives now. It is in your phone, your inbox, your search results, and probably on at least one screen at the front desk. And yet, how many times have you asked it something and found the reply frustratingly off? You ask for a short note and get four paragraphs. You ask a simple question and get a lecture with headings. It agrees with everything you say, invents a detail you never gave it, and halfway through forgets what you told it two messages ago.
If that sounds familiar, you are using it the way most people do. Open the app, type a sentence, hope for the best. Most AI prompts look exactly like that, and the reply is only ever as good as what you gave it. We give a new receptionist a proper briefing on day one. We give the AI eleven words.
This post is about the small changes that fix that. Nothing technical and no new software. A habit that takes twenty seconds, a dozen AI prompts you can copy as they are, and a separate list of things AI can take off your plate that have nothing to do with email.
Why the answers come out wrong
Better AI answers start with one fact nobody mentions when you sign up: the AI knows nothing about you. It does not know you work in a medical office, that most of your patients are over 65, that the notice has to fit on a half sheet, or that your manager removes exclamation marks on sight. Every new chat starts blank.
So when you type “write a notice about our new hours,” it fills every gap with the most average guess it has. Average tone, average length, average reader. That is not the AI being dim. It is the AI doing exactly what you would do if someone handed you that sentence and walked away.
The other habit that costs people is treating the first reply as the finished product. You read it, sigh, and close the window. It was a draft. “Shorter,” “less formal,” or “you left out the parking” takes four seconds to type, and the second version is almost always the one you wanted. The people getting real use out of AI are not better at writing AI prompts. They just reply.
One habit that fixes most AI prompts
Before you hit enter, tell it what you would tell someone new on their first morning: who you are, who this is for, what you need, how long it should be, what it should sound like, and what to leave out. That is the whole trick. It takes about twenty seconds.
Side by side, it looks like this. “Write a holiday closure notice” gets you three cheerful paragraphs you would never post. “I manage a family practice in Ontario. Write a waiting-room notice for patients, under 80 words, warm but plain, saying we are closed December 24 to 27 and that urgent concerns should go to 811 or the nearest urgent care. No exclamation marks.” That one comes back ready to print.
Two more things help. Paste in something you already wrote, a past notice or an email, and say “match this.” It copies your voice far better than it invents one. And the same ChatGPT prompts and follow-ups work in Copilot and Gemini, so there is nothing to relearn if your office ever switches tools.
12 AI prompts that get better answers
Copy these AI prompts as written and add your own details where it says so. Each one fixes something that goes wrong a lot.
- “Before you answer, ask me the three questions that would most improve your response.” Stops the AI from guessing and makes you fill in the gaps you forgot.
- “Only use the information I give you. If something is missing, say so instead of filling it in.” The single best defence against made-up facts.
- “List the assumptions you made in that answer.” You will be surprised how often it assumed the wrong province, the wrong audience, or the wrong year.
- “Here is my draft. Tighten it without changing the meaning or the tone.” Gets you an edit instead of a rewrite.
- “Match the style of this sample.” Paste in something you wrote yourself. This is the fastest way to make AI sound like you.
- “Give me three versions: one short, one standard, one formal.” Choosing is quicker than tweaking.
- “Rewrite this at a grade 6 reading level without losing any of the instructions.” Essential for anything a patient will read.
- “Read this as a first-time patient filling out our intake form. What is confusing or missing?” Turns the AI into a test reader instead of a writer.
- “What would a skeptical reader push back on?” Finds the weak spot in a letter, a proposal or a policy before someone else does.
- “Explain this error message in plain language and tell me whether it is safe to click OK or whether I should call IT.” Saves a ticket, or tells you to open one.
- “Turn these notes into action items with an owner and a due date for each.” Paste the messy notes exactly as they are.
- “Summarize this in ten bullet points and flag anything about auto-renewal, price increases or cancellation notice periods.” For vendor contracts and the long emails you have been avoiding.
A tip for the last two: the more you paste, the better the answer, as long as what you paste is safe to share. More on that below.
Beyond email: 12 other ways to use AI at work
Most people stop at “help me write this email,” which is a bit like buying a truck to carry groceries. If you have been wondering how to use AI at work beyond the inbox, this is the list. None of it needs a patient’s name to work.
- Read what you do not have time to read. Paste a long vendor email, a 30-page policy update or a new guideline and ask for the five things that affect your office.
- Get ready for a meeting. Turn a messy list of concerns into an agenda, or ask “what should I ask this vendor before we sign?”
- Fix the spreadsheet. Ask for the formula you need, or paste the one that is broken and ask why it returns an error.
- Make your blank forms and handouts readable. The template, not a patient’s record. Paste the intake form or the after-visit sheet as it sits on the shelf, ask for a grade 6 reading level, and watch the follow-up calls drop.
- Build staff training. A five-question phishing quiz, an onboarding checklist for a new receptionist, a one-page how-to for the phone system.
- Hire faster. A job posting that sounds like your office, a set of interview questions, and a simple scoring sheet so every candidate gets the same ones.
- Write front-desk scripts. Appointment confirmations, no-show follow-ups, and a calm response for the caller who is already upset.
- Rehearse the hard conversation. Ask it to play an upset patient or a pushy salesperson, no names and no real details, and practise before the real call.
- Decode jargon. A cyber insurance questionnaire, a lease clause, a line in a privacy regulation. Get the plain-language version first, then confirm with whoever signs off.
- Ten ways to reduce no-shows. Five ideas for a long wait in the waiting room. Names for a new wellness program.
- A holiday coverage schedule, a checklist for an office move, a timeline for replacing aging computers.
- Draft a translation. A waiting-room notice in a second language, checked by a fluent speaker before it goes on the wall. Never for clinical instructions.
Every one of these works without a single patient’s name in it, and that is the point.
Beyond email: 12 other ways to use AI at work
Most people stop at “help me write this email,” which is a bit like buying a truck to carry groceries. If you have been wondering how to use AI at work beyond the inbox, this is the list. None of it needs a patient’s name to work.
- Read what you do not have time to read. Paste a long vendor email, a 30-page policy update or a new guideline and ask for the five things that affect your office.
- Get ready for a meeting. Turn a messy list of concerns into an agenda, or ask “what should I ask this vendor before we sign?”
- Fix the spreadsheet. Ask for the formula you need, or paste the one that is broken and ask why it returns an error.
- Make your blank forms and handouts readable. The template, not a patient’s record. Paste the intake form or the after-visit sheet as it sits on the shelf, ask for a grade 6 reading level, and watch the follow-up calls drop.
- Build staff training. A five-question phishing quiz, an onboarding checklist for a new receptionist, a one-page how-to for the phone system.
- Hire faster. A job posting that sounds like your office, a set of interview questions, and a simple scoring sheet so every candidate gets the same ones.
- Write front-desk scripts. Appointment confirmations, no-show follow-ups, and a calm response for the caller who is already upset.
- Rehearse the hard conversation. Ask it to play an upset patient or a pushy salesperson, no names and no real details, and practise before the real call.
- Decode jargon. A cyber insurance questionnaire, a lease clause, a line in a privacy regulation. Get the plain-language version first, then confirm with whoever signs off.
- Ten ways to reduce no-shows. Five ideas for a long wait in the waiting room. Names for a new wellness program.
- A holiday coverage schedule, a checklist for an office move, a timeline for replacing aging computers.
- Draft a translation. A waiting-room notice in a second language, checked by a fluent speaker before it goes on the wall. Never for clinical instructions.
Every one of these works without a single patient’s name in it, and that is the point.
One rule that never bends
AI in healthcare comes with one line no prompt is worth crossing. A patient’s name, health card number, date of birth, or anything else that could identify them NEVER goes into public AI, including the free versions most people use. The Office of the Privacy Commissioner of Canada puts it simply: what you type in may be collected, stored and used to train the system.
Business versions with data protection are safer, but the habit is the same. Strip out anything identifying before you paste, and treat every answer as a draft a person checks before it leaves the office. A simple test: if you would not post it on the clinic’s website, do not paste it into AI.
Start with two
You do not need all 24 ideas on this page. Pick two AI prompts from the first list and one idea from the second, try them this week, and notice how much less you are arguing with the screen.
If you are a BlueBird iT client and not sure whether the AI your office uses is the safe kind, ask your Client Services Representative. It is a five-minute conversation, and we would rather have it earl
