At 9:40 on a Thursday night, a tenant on the third floor gets a text. It says the water will be shut off “for maintenance” tomorrow. It doesn’t say for how long. It doesn’t say why. It doesn’t say who to call. She has a baby, and she starts wondering how she’ll make bottles in the morning.

The next morning, the building’s office gets eleven calls. Several are angry. One tenant has already filled the bathtub, just in case. The maintenance crew arrives and finishes in ninety minutes. Nobody needed to worry. The message just didn’t say enough, and the way it was written made people feel like an afterthought.

Now picture the same night with a different message. It says the water will be off from 9 to 11 a.m. to replace a valve that’s been leaking in the basement. It apologizes for the inconvenience. It suggests filling a pitcher tonight. It gives the office number and says someone will answer. The tenant reads it, fills a pitcher, and goes to bed.

The difference between those two messages isn’t technology. It’s care and clarity. But AI tools can help a small team deliver care and clarity more consistently, if a person stays in charge of what goes out. This piece follows an invented property management company through that process. The company, its staff, and every example are illustrative, not a client story, and nothing here is a reported result.

Why tenant messages go wrong

Property managers write a lot. Maintenance notices. Rent reminders. Lease renewal letters. Rules about parking, trash, and packages. Replies to complaints. Updates after a storm. Each message is short, and each is written in a hurry, often by whoever is at the desk.

That’s where the problems start:

• Messages are written for the sender, not the reader. “Scheduled maintenance on the domestic water system” means something to the maintenance supervisor. It means very little to a tenant.
• Key facts get left out. When, how long, what to do, and who to call.
• The tone drifts. One manager is warm; another is curt. A tenant who gets both starts to wonder which one reflects how the company really feels.
• Hard messages get harder. Late-rent notices, lease violations, and entry notices can sound threatening even when they don’t need to.
• Nobody has time to edit. The message that goes out is the first draft.

None of this means the staff doesn’t care. It means writing well under pressure is hard.

What plain language actually means

Plain language isn’t dumbing things down. It’s writing so the reader can find what they need, understand it the first time, and act on it.

The federal government has taken this seriously for years. The Plain Writing Act of 2010 requires federal agencies to use clear communication the public can understand. Digital.gov keeps a plain language guide series that explains the core idea: write for your specific audience, and organize content around what they need to know.

That guidance was written for government, but the principle applies anywhere. A tenant reading a notice at 9:40 p.m. is an audience with a specific need. They want to know what’s happening, how it affects them, and what to do.

Clear is also accessible

Clear writing also helps people with disabilities, people reading in a second language, and people who are stressed or tired. The W3C’s Web Accessibility Initiative offers tips for writing accessible web content, and many of them translate directly to tenant messages: provide clear instructions, keep content clear and concise, use headings to convey structure, and make link text meaningful instead of writing “click here.”

For a property manager, that might look like:

• Putting the most important fact first: “Water off Friday, 9 to 11 a.m.”
• Using short sentences and common words.
• Giving one clear action: “Fill a pitcher tonight if you’ll need water in the morning.”
• Writing links that say what they open: “See the full maintenance schedule” instead of “click here.”
• Avoiding messages that rely only on an image or a scanned PDF with no text.

Where AI helps

AI writing tools are good at some of the work that slows a busy office down. Used carefully, they can:

Turn shorthand into a first draft. A maintenance supervisor types, “water off fri 9-11 replacing basement valve.” The tool turns that into a complete, readable notice. A person still reviews it.

Rewrite for tone. A manager drafts a late-rent reminder that sounds sharp. The tool offers a version that’s still clear about the facts and deadline but sounds like it came from a person who’d like to help.

Check for missing pieces. A draft notice can be compared against a simple checklist: what, when, how long, why, what to do, who to call.

Adapt length to the channel. A text message needs to be short. An email can include more detail. A posted notice in the lobby can include a little more context.

Offer versions in other languages. For buildings with tenants who read other languages, an AI draft can be a starting point. A fluent person should review any translation before it goes out, especially for anything about money, safety, or legal rights.

Where AI stops

AI doesn’t know your tenants. It doesn’t know that the woman on the third floor has a baby, or that the tenant in 2B has been dealing with a leak for weeks and is already frustrated. It doesn’t know your lease, your local rules, or what your owner has approved. It can produce a confident, polished message that’s wrong.

So the company draws clear lines.

AI never sends anything on its own about money, legal rights, or lease terms. Late fees, rent increases, lease violations, notices to enter a unit, and anything that might be part of a legal process are drafted by a person or reviewed line by line by a person before they go out. Rules for notices vary by state and city. This is general information, not legal advice, and a property manager should rely on their attorney and local requirements.

AI doesn’t make promises. If a draft says “the repair will be finished by noon,” a person confirms that’s true before it’s sent.

AI doesn’t respond to complaints without review. A tenant who writes in upset deserves a reply from a person, even if an AI tool helped draft it.

AI doesn’t treat tenants differently. Messages about rules, fees, and policies should be consistent for every tenant. A manager reviews drafts with fairness in mind.

Naming the human step

Every AI workflow in this company has a named human step. It’s written down, and it’s someone’s job.

For the property management company, the steps look like this:

1. The maintenance supervisor or office manager writes the facts. What’s happening, when, how long, and why.
2. An AI tool turns the facts into a draft, following the company’s written tone and checklist.
3. The office manager reads the draft out loud, or at least slowly, and fixes anything that’s unclear, wrong, or cold.
4. The office manager approves and sends it, or holds it for the property manager if it touches money or lease terms.
5. Replies from tenants go to a person, who answers or decides whether the tool can help draft the answer.

The human step isn’t a rubber stamp. If the office manager is too busy to read the draft, the message waits or gets written by hand.

Writing the company’s tone down

AI tools work better with clear instructions. So does a team. Before turning anything on, the company writes a short tone guide. It’s one page. It includes:

• Who we are to tenants: neighbors who take care of the building, not landlords who issue orders.
• How we sound: warm, direct, calm. We apologize when something is inconvenient. We don’t over-apologize or make excuses.
• What every notice includes: what, when, how long, why, what to do, who to call.
• Words we avoid: jargon like “domestic water system,” vague words like “soon,” and threatening phrases unless a legal notice requires specific language.
• How we handle hard messages: state the facts plainly, explain what happens next, and offer a way to talk.

That guide becomes the instruction set for any AI tool and the training document for new staff. Same rules for both.

What this looks like with an AI receptionist

Some property management companies also use an AI receptionist on their website chat or phone line to answer common questions after hours. Depending on your edition and how your account is set up, a platform like the one WeKinnect works with can give that receptionist a written role and tone, answer from a knowledge base you provide (such as office hours, how to submit a maintenance request, and parking rules), and use different instructions for different channels, such as short plain text for text messages and longer replies by email.

A few things matter here:

• The receptionist should say when it doesn’t know. Depending on your setup, it can tell the person it doesn’t have enough information and that a team member will follow up. That’s far better than a confident wrong answer.
• Emergencies need a person. A tenant reporting a gas smell, a flood, or a lockout should be routed to a human right away, with clear instructions, such as calling 911 for a gas emergency.
• The knowledge base has to stay current. If office hours change or the maintenance request process moves to a new form, the receptionist will keep giving the old answer until someone updates it.
• Conversations get reviewed. Depending on your setup, conversation logs can be reviewed so the team can see which questions the receptionist handled poorly and fix the instructions. A weekly review is a sensible starting point.

None of this is a promise of fewer calls or happier tenants. It’s a tool that can help a small office answer simple questions consistently, if someone keeps it honest.

Audit before you turn anything on

Before using AI to draft tenant messages, the company does a short audit. It takes an afternoon.

Gather recent messages. Pull twenty or thirty notices, reminders, and replies from the past few months.

Sort them. Which were routine? Which were sensitive? Which caused confusion or complaints?

Mark the lines. Decide which kinds of messages AI can help draft, which need line-by-line human review, and which should never involve AI.

Write the checklist and tone guide. Use what you learned from the messages that worked and the ones that didn’t.

Test with past examples. Give the tool the facts from an old notice and compare its draft to what was actually sent. Is it clearer? Is it accurate? Does it sound like the company?

Decide who approves. Name the person for each kind of message, and name a backup.

Only then does the company start using AI for live messages, beginning with routine maintenance notices.

A few examples, before and after

Here’s how the company’s routine messages might change. These are illustrative examples, not real notices.

Rent reminder, before: “Rent is due on the 1st. Late fees apply after the 5th.”

Rent reminder, after: “A friendly reminder that rent for next month is due on the 1st. If you’re having trouble paying on time, please call the office before the 5th. We’d rather talk with you early.”

The second version keeps the facts. It just sounds like it came from a person. Any mention of late fees still follows the lease and local rules, reviewed by the property manager.

Pest treatment, before: “Pest control will service all units Wednesday.”

Pest treatment, after: “Pest control will treat all units on Wednesday between 10 a.m. and 2 p.m. Please clear the area under your kitchen sink and keep pets in a closed room or carrier during the visit. Questions? Call the office, and we’ll help.”

Parking, before: “Unauthorized vehicles will be towed.”

Parking, after: “Starting Monday, the lot will be repainted, and we’ll need it empty from 8 a.m. to 5 p.m. Street parking is available on Elm and Fourth. Vehicles left in the lot may need to be moved, and we’d much rather not tow anyone. Thanks for your help.”

In each case, a person wrote or confirmed the facts, and a person approved the final message.

When something goes wrong

The hardest messages come after a problem: a burst pipe, a heating failure in January, a break-in in the parking garage, a power outage that lasts longer than anyone expected. Tenants are anxious, and the office is overwhelmed.

These are the moments when a clear, kind message matters most, and when an AI tool can help a person get the first draft down fast. But they’re also the moments when a wrong detail does the most damage.

The company’s rule for incidents is simple: a person writes the facts, confirms them with whoever is on site, and approves every update. AI can help structure the message and keep the tone calm. It never guesses at a restoration time, a cause, or what the company will pay for.

A good incident update follows a pattern:

• What happened, in one sentence.
• What it means for tenants right now.
• What the company is doing about it.
• When the next update will come, even if there’s nothing new.
• Who to call, and what to do in an emergency.

That fourth point is easy to forget. “We’ll send another update by 6 p.m., even if the work isn’t finished” tells tenants they won’t be left wondering. Then someone has to actually send it.

The follow-up nobody writes

Most offices send a notice before maintenance. Fewer send one after. A short follow-up, such as “The water is back on. Thanks for your patience. If anything isn’t working right, let us know,” closes the loop. It also catches problems early, like a tenant whose faucet is now sputtering.

This is a good place for AI help. The follow-up is routine, low-risk, and easy to forget. A draft can be prepared alongside the original notice, held until the work is confirmed finished, and sent by a person once it is.

Building a library of approved messages

Over time, the company builds a small library of messages it has already reviewed and approved: the standard water shutoff notice, the pest treatment notice, the move-in welcome, the holiday office hours. Each one has blanks for the specifics.

The library does two things. It gives AI tools strong examples to follow, and it gives new staff a starting point that already sounds like the company. When a message in the library causes confusion, the team fixes it once, and every future version improves.

How the company will know it’s working

The company doesn’t expect AI to produce a particular number of fewer calls or complaints, and it won’t promise anyone that. Instead, it watches a few simple signs over a few months. Are tenants calling to ask questions the notice should have answered? Are complaints about tone going down or staying the same? Are staff spending less time rewriting messages? Those are honest signals, checked by a person, and they tell the team where to adjust.

Common mistakes with AI and tenant communication

Letting AI send messages without review. The tool is a drafting assistant, not a manager.

Feeding it incomplete facts. If the input says “maintenance tomorrow,” the draft will guess at the rest. Give it the facts.

Assuming polite equals clear. A warm message that leaves out the time window still fails the tenant.

Using AI for legal notices. Formal notices often have required content and delivery rules. Those belong to the property manager and the attorney.

Forgetting to update the instructions. When the office moves, the phone number changes, or policies change, update the tone guide, the checklist, and any knowledge base.

Expecting AI to fix a relationship problem. If tenants are frustrated because repairs take too long, a nicer message won’t solve it. Fix the repair process too.

How a youth sports league would use this differently

The same thinking works in a very different setting. Imagine a volunteer-run youth sports league with a few hundred families.

Its messages are things like weather cancellations, schedule changes, uniform pickup, and volunteer reminders. Like the property management company, it writes in a hurry, often late at night, and the volunteers who write messages change every season.

An AI tool could help the league turn quick notes into clear messages, keep the tone friendly and consistent, and adapt messages for text and email. The human step would be a named volunteer, such as the league secretary, who reviews every message before it goes out.

The lines would be different, though. The league would keep anything about a child’s safety, an injury, a conduct issue, or a dispute between families strictly human. It would never let an AI tool make decisions about which team a child plays on or whether a game is canceled. Those choices belong to people, and the families need to know a person made them.

Kindness is a decision, not a feature

AI tools can help a busy team write clearer, warmer messages. They can’t decide to be kind. That decision belongs to the people running the business: the tone they choose, the facts they insist on, and the review step they refuse to skip.

The tenant on the third floor doesn’t care what tool wrote the message. She cares that it told her what she needed, at a time she could use it, in a voice that sounded like someone was looking out for her.

If you’d like help deciding where AI can support your team’s messages and where a person should stay in charge, start a conversation with WeKinnect. Bring a few recent messages, and we’ll look at them together.

Photo by Patrick Tomasso on Unsplash