In early December, a man in his sixties gets a text from the physical therapy clinic where he rehabbed his knee last spring. It says the clinic will be closed the week between Christmas and New Year’s. Then, ten minutes later, he gets another one: a reminder that he “still has visits available this year” and should “book now before benefits reset.” He’s been fine for months. He wonders whether the clinic thinks something is wrong with him, or whether it’s just trying to fill the schedule.
Down the street, a different clinic sends a different message to its former patients. It’s a short note, written by the therapist who treated them, thanking them for the work they put in and asking how they’re doing. It mentions the holiday hours at the bottom. It doesn’t ask anyone to book anything. A few people reply with updates. One mentions her shoulder has been bothering her again, and the front desk calls her to see whether she’d like to come in.
Both clinics used the same tools. The difference was the plan behind the messages, and who decided what they said.
This piece follows an invented physical therapy clinic as it plans its year-end outreach with help from AI. The clinic, its people, and its patients are illustrative, not a client story, and nothing here describes a real result. Health care communication has specific privacy rules. Where we mention them, it’s general information, not legal advice.
The end of the year is crowded. Every business sends messages: holiday hours, thank-yous, last-chance offers, year-in-review emails. In health care, there’s an extra pressure. Many patients’ insurance benefits or deductibles reset in January, and clinics know that some patients would benefit from finishing care before then.
That’s where good intentions can turn into messages that feel pushy or confusing:
AI can make these problems worse, because it makes it easy to send more messages faster. Used carefully, it can also make them better.
It’s easy to talk yourself out of a thank-you note. Dana’s first thought, when her office manager suggested personal notes to former patients, was that people wouldn’t care. They’d finished their care. They had moved on.
Research suggests that instinct may be off. In a set of experiments published in the Journal of Experimental Psychology: General, Amit Kumar and Nicholas Epley found that people performing small acts of kindness consistently underestimated how positive the recipients would feel. Recipients’ reactions were lifted by the warmth the act conveyed, while the people doing the kind thing paid less attention to that warmth. The researchers suggest this miscalibration can keep people from doing kind things more often.
The studies were about everyday kindness, like handing a stranger a cup of hot chocolate, not about clinics or outreach, so they don’t predict any business result. But they’re a useful nudge. A two-sentence note from the therapist who helped someone through a hard recovery may mean more to them than the therapist expects. That’s a reason to write it, and a reason for a person, not a template, to sign it.
Before anyone opens an AI tool, the clinic’s owner, a physical therapist we’ll call Dana, sits down with her office manager and lead therapist to plan. They answer four questions.
Who are we reaching out to? They list the groups: current patients, patients who finished care this year, patients who stopped coming before their plan was done, referring physicians and their staff, and the clinic’s own team.
What does each group actually need from us at the end of the year? Current patients need holiday hours and schedule changes. Patients who finished care might appreciate a thank-you and a check-in. Patients who stopped early might need a gentle, no-pressure invitation to reconnect. Referring offices deserve thanks. The team deserves thanks most of all.
What should each group not receive? Patients who finished care successfully shouldn’t get booking pressure. Nobody should get more than a couple of messages in December.
Who decides what goes out? A person reviews every message before it’s sent.
That plan comes first. AI comes second.
With a plan in hand, AI tools can save the small team real time.
Drafting from a few facts. The office manager types the holiday hours and a few notes about tone, and an AI writing tool produces a clear, friendly draft. She edits it.
Adjusting tone. Dana writes a quick, slightly clinical thank-you. The tool suggests a warmer version that still sounds like her. She picks what she likes.
Adapting for channels. A text needs to be short. An email can be longer. A note card for referring physicians can be more personal. The tool helps produce versions for each, and a person approves each one.
Checking for missing pieces. A draft about holiday hours can be checked against a simple list: which days, what hours, how to reach someone in an urgent situation, and when normal hours resume.
Answering common questions after hours. Depending on your edition and how your account is set up, the WeKinnect Marketing Platform’s AI receptionist can answer routine questions on your website chat or phone line, such as holiday hours and how to request an appointment, using a knowledge base you provide. If it doesn’t have the answer, it can say so and collect contact details so a team member can follow up. That keeps the front desk from fielding the same holiday-hours question fifty times.
AI doesn’t know these patients. It doesn’t know that one man’s knee is fine and he’d be annoyed by a booking reminder, or that another patient stopped coming because of a family emergency. It can write a warm, confident message that’s wrong for the person receiving it.
So the clinic draws clear lines:
AI doesn’t decide who gets which message. People decide the groups, based on what the clinic knows. AI can help draft, but the list is a human decision.
AI doesn’t write anything clinical. Messages never mention a patient’s condition, treatment, or progress. Anything about care belongs in a conversation with a therapist, not in a mass message.
AI doesn’t create urgency. No “last chance,” “act now,” or “don’t lose your benefits” language. If a patient might benefit from finishing care, a therapist reaches out personally.
AI doesn’t send on its own. Every message is reviewed by a person before it goes out.
Every message in the clinic’s year-end plan has a named human step.
If anyone is too busy to do their step, the message waits.
Here’s an illustrative example of a thank-you note to a former patient. It’s an invented example to show tone, not a real message.
“Hi Ellen, it’s Priya from the clinic. I was thinking about the folks I worked with this year, and you came to mind. You put in real work, and I hope you’re back to doing the things you love. Thank you for trusting us. If anything comes up, you know where to find us. Wishing you a warm holiday season. Our holiday hours are below, just so you have them.”
Notice what it doesn’t say. It doesn’t name Ellen’s injury. It doesn’t ask her to book. It doesn’t mention her insurance. It sounds like Priya, because Priya wrote it, or at least rewrote the AI draft until it sounded like her.
Here’s the version the clinic decided not to send, also illustrative:
“Dear Valued Patient, as the year comes to a close, don’t forget to use your remaining visits before your benefits reset! Book today to stay on track with your recovery.”
It isn’t cruel. It’s just not about Ellen. It’s about the schedule.
The holiday weeks are when handoffs fail. Staff are out, hours change, and messages pile up. The clinic plans its handoffs before December starts.
Depending on your edition and how your account is set up, the platform’s voice receptionist can answer calls from the knowledge base, say when it can’t find an answer, offer to take a message and have someone call back, and transfer calls to set numbers based on what the caller needs and conditions like time of day. After-hours behavior can be configured to transfer or take a message. Which of these are available depends on the Conversations AI edition, so the clinic confirms that first.
The clinic decides:
Before the holiday hours take effect, the office manager tests the receptionist. Depending on your setup, configurations can be saved and tested with simulated conversations before they go live. She asks the questions patients actually ask: Are you open the day after Christmas? Can I come in on New Year’s Eve? My back hurts, what should I do? She checks that each answer is accurate and that the pain question goes straight to a person.
Once it’s live, the team reviews conversation logs to find questions the AI couldn’t answer well. The vendor’s guidance suggests weekly as a good starting point, changing one thing at a time. During the holiday weeks, the clinic checks more often. An “Explanation” on any AI response shows which knowledge sources it used, which helps the team find outdated information quickly.
A health care clinic has to think about privacy in a way many businesses don’t. Federal rules under HIPAA govern how protected health information can be used and shared, and states may add their own requirements.
The U.S. Department of Health and Human Services explains the HIPAA Privacy Rule’s approach to marketing communications. In general terms, it describes marketing as a communication about a product or service that encourages people to buy or use it, and says that, with limited exceptions, using protected health information for marketing requires the person’s written authorization. It also describes communications that aren’t considered marketing, such as communications for a person’s treatment, like refill reminders, and communications describing a provider’s own health-related services.
This is general information, not legal advice, and the details matter. A clinic should check its year-end outreach plan with its compliance advisor before sending anything that uses patient information. But the principle is a helpful guide for kindness, too. A message that’s genuinely about someone’s care, sent by the people who provided it, feels very different from a message that uses their information to sell them something.
The clinic adds a few practical rules:
The clinic plans a simple December calendar, so no one gets too many messages:
Early December: holiday hours for current patients. One message, by each patient’s preferred channel, with the closure dates and how to reach someone if they have an urgent question. The AI receptionist is updated with the same hours.
Mid-December: thank-you notes to former patients. A short, personal note from the therapist who treated them, for patients who finished care this year. No booking request. Holiday hours appear at the bottom only as information.
Mid-December: a gentle note to patients who stopped early. Written carefully, reviewed by Dana, and sent only to people the therapists agree it’s appropriate for. It says the clinic is thinking of them and is there if they’d like to talk, with no pressure and no mention of benefits.
Mid-December: handwritten cards to referring offices. Not automated. Dana writes them herself, with help from AI only for a first draft if she wants it.
Late December: thanks to the team. In person, at the clinic’s year-end lunch.
That’s it. Each group hears from the clinic once, maybe twice, and every message has a purpose.
Before using AI for any of this, the clinic writes a one-page tone guide. It says:
That guide becomes the instructions for the AI tools and the reference for the team. Depending on your edition and how your account is set up, the platform’s AI tools can be given a written role and tone, and different instructions for different channels, such as short plain text for texts and longer replies by email.
Before the first year-end message goes out, the clinic audits its plan:
The clinic doesn’t set targets for bookings or replies, and it doesn’t promise itself any particular result. Instead, it watches a few honest signs in January:
Those answers shape next year’s plan.
Sending everything to everyone. Different groups need different messages, or none at all.
Letting urgency creep in. It’s easy to write “before the year ends.” Ask whether the deadline matters to the patient or only to the schedule.
Automating thanks. AI can help draft a thank-you, but a person should write or approve it and sign it.
Forgetting the team. The people who did the work all year deserve the first thank-you.
Ignoring replies. If you invite people to respond, someone has to read and answer.
The same thinking applies to a very different organization. Imagine a volunteer community choir planning its year-end messages.
Its groups might be singers, audience members who came to the winter concert, donors, and the church or school that lends rehearsal space. An AI tool could help draft a thank-you to singers, a concert-recap note to audience members, and a short letter to the host venue.
Its lines would be different, but just as clear. A person would decide who gets each message. The music director would write or approve the note to singers. Donor thank-yous would be accurate and personal. Nobody would get a fundraising appeal disguised as a thank-you. And with no health information involved, privacy would still matter: no sharing singers’ contact information, and no posting photos without permission.
The man in the opening didn’t need two texts in ten minutes. He would have welcomed a short note from the therapist who helped him walk without pain again.
That’s the real lesson of year-end outreach. Kindness isn’t a feature you turn on. It’s a plan: who you’re reaching out to, what each person needs, what they shouldn’t receive, and who checks every message before it goes. AI can help with the writing. People decide what’s kind.
If you’d like help planning year-end outreach that feels personal and respectful, start a conversation with WeKinnect. Bring last December’s messages, and we’ll look at them together.
Photo by Judy Beth Morris on Unsplash
Written By
Maren Hale focuses on brand positioning at WeKinnect Global Branding Agency. Guided by the core value of Service before Self, she writes for entrepreneurs and family-owned legacy businesses. Maren writes on Business Listings, Google Business Profile, SEO, GEO (being found by AI tools like ChatGPT, Gemini, and Grok), and Community. WeKinnect serves businesses across restaurants, arts, travel and tourism, banking, home services, nonprofits, and the pet industry. Maren brings that same spirit of service to every piece she writes.
Darlene Gagnon
16 Oct 2026
