“AI agents will run your entire property management company soon!” I’m sure you’ve seen this kind of claim on LinkedIn or from an industry newsletter. If you’re a regular reader of this publication, you know that I’m a big believer in AI, but I’m also a big skeptic of a lot of these claims.
The first thing we need to remember is that this shouldn’t even be the goal, even if it was actually possible (it’s currently not). If it was true that everything involved in property management could just be automated, then all of us would quickly be out of a job because some SaaS company would replace the entire industry with an AI product that any landlord could sign up for for $9.99/mo. But you and I both know that property management is a lot more than just automated back office tasks. It’s a people business. And people don’t like robots.
PMAssist Industry Insights is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
The rapid rise of an open source software platform called “OpenClaw” has accelerated the rhetoric on this topic. You’ve probably seen people writing about it, myself included. I gave OpenClaw a try, and to be blunt, I don’t think any PM company should be operating on OpenClaw. We’ll dive into the various reasons why in this article, but I do want to emphasize that I absolutely do support the idea of using AI to help automate portions of your operation. I just don’t believe that a glorified digital internet with unlimited curiosity is the answer for that. A much more dependable system with agents with specific roles tightly programmed is a much safer and more pragmatic approach.
What is OpenClaw, and why are people excited about it?
OpenClaw describes itself as a self-hosted AI assistant for “genuine automation.” But what does that really mean?
Basically, OpenClaw is just a software program that uses an AI large language model (LLM) of your choosing and a communications method of your choosing to talk with it (such as Slack or Telegram) to perform automation tasks. It is “open source,” which means that literally tens of thousands of developers are freely devoting their time to building it, and it is freely available to anyone who wants it. That obviously makes it quite attractive to anyone looking to play around in the agentic AI sandbox, but the open nature of it also creates some unique issues.
OpenClaw supports a lot of integrations and “skills” that you can just add on at will, and you can set up multiple OpenClaw agents to talk to each other and even have one “brain” agent to coordinate all of these agents together. I must say, it’s really quite something when you build a few of these agents and then watch them start talking to each other in Slack. Yes, a little bit dystopian, but also quite impressive. It’s easy to see why so many people immediately latch on to it.
Stay focused on your properties, and let Enterprise Bank & Trust handle your banking. Property management companies from across the country partner with Enterprise for:
Competitive analysis program to offset bank fees and third-party invoices
Property management trust account experts with dedicated support teams to help you open and manage your accounts
Streamlined loan process for simple or complex lending needs
Watch to learn more about our tailored solutions for property managers and their accounts:
“Holding your trust account with the property management team at Enterprise is a no-brainer. Allison understands all of the regulations involved with property management trust accounts. When I transferred my trust accounts over, the team was very involved and made it a seamless process. The credit system they provide is awesome as well. This banking team is hands-down the best option for trust accounts in the property management industry.”
Mike Catalano General Partner and Co-Founder PURE Property Management
It’s important to note what the developers of OpenClaw were and are actually trying to accomplish with it. OpenClaw isn’t really intended to be a basic automation platform that completes rote administrative tasks, or even basic reasoning. OpenClaw is designed to be an agent that can solve open-ended reasoning problems and act autonomously to solve them. For example, let’s say you’re planning a trip to Europe in the fall, and you don’t want to do all of the work to make the arrangements and plan the trip. OpenClaw is perfect for being that kind of digital personal assistant (although there are security problems I’ll get to with even that). It’s an open-ended project that has a million possible permutations. You could see the Eiffel Tower, or you could visit the Vatican. You could take a train, or you could rent a car. The possibilities are endless, and a broad reasoning model is good for handling something with lots of possibilities.
But that’s simply not what property management is. At least not 95% of the time. Sure, there are very unusual situations that pop up from time to time, but I doubt in our highly regulated and highly litigious industry that those are the things you want agentic AI handling for you. That’s where you actually do want to be personally involved, and you want the agentic AI doing all of the day-to-day busy work instead so that you can actually focus on those open-ended problems.
Most useful PM workflow automations are quite narrow and predictable. Those who have been in the industry for a while probably remember a PM out of Texas named Steve Crossland. Steve retired years ago now and enjoys most of his time in various beautiful Mexican locales, but he used to be on the PM Facebook groups all the time and he had a common saying: “property management is just a series of binary choices.” In other words, it’s the 4th of the month and the rent has either been paid in full or it hasn’t been; if it hasn’t, send out the cure notice. Property management is essentially thousands of those little binary paths stacked on top of one another. And while that is something that AI is fantastic at handling, that’s not what OpenClaw is designed to handle.
One of the problems I see growing around the AI issue is what I’ll call the “agent-shaped hammer problem.” We’ve all heard the saying “if all you have is a hammer, everything starts to look like a nail.” Similarly, once you’ve heard the AI gurus talking nonstop about agentic AI and you’ve gotten a little taste of it, suddenly everything starts looking like an AI reasoning problem, even when it’s really nothing more than a very simple routing or rules automation problem. In plain language, you don’t need OpenClaw when all you need to happen is a binary choice based on the circumstances.
Security Is Not a Side Issue Here; It is THE Issue
If any system touches resident & owner communications, financial data, contract documents, accounting, or anything similar, it needs to be absolutely boring and locked down. This is not a business where experimenting with open source software on live data is a realistic option.
The bottom line is that OpenClaw’s power is also its biggest risk. It can connect across all of your various tools autonomously, it can execute actions based on its own reasoning (or at least what we’re calling “reasoning” in this pre-AGI environment), and it can fully control your systems. This is obviously incredibly powerful, and it’s easy to see why it’s compelling. But this should also be an enormous red flag for PMs who are entrusted with very sensitive data and who exist in a very risky industry.
Recent events make this a lot more than just a hypothetical problem. Oasis Security, a leading AI security agency, recently disclosed what’s called “ClawJacked,” a vulnerability that allowed malicious websites to hijack a locally running OpenClaw machine. All you had to do was accidentally visit the malicious website, which could easily be disguised as a great new plugin for OpenClaw, and then ClawJacked would establish a connection to your local computer running OpenClaw, brute-force its way into it by jamming through hundreds of password guesses per second, and then it would be able to locally control your OpenClaw system and do whatever it wants with it. So if you had connected your OpenClaw to your property management accounting system, for example, suddenly it could gain access to all of that financial and personal data. And if you gave OpenClaw access to your email (something almost everyone does who uses it), then it could act as you, hide any responses to messages without your knowledge, etc.
And this is just the tip of the iceberg. In only February of this year, OpenClaw published numerous security advisories covering all sorts of vulnerabilities in the system. Granted, many of these issues, ClawJacked included, have been addressed by patches already, but this just shows you how easy it is for malicious actors to cause all sorts of havoc when you give an AI this much latitude.
For a PM company to give AI this much power would require a payoff that is absolutely overwhelming to make the risk worthwhile. And frankly, the payoff just isn’t there. You can accomplish what you want to accomplish without this sort of autonomous AI.
Remember, when your software is designed to do more than you need it to do, then it also creates more ways for it do the wrong thing.
Property Management AI Needs Accountability, Not Autonomy
When something bad happens in property management, you need to quickly be able to figure out what happened, why it happened, what data input led to it, and who approved it. This is all very easy with deterministic automation workflows (binary choices). It is all very complex and nearly impossible when systems are autonomous black boxes.
Essentially, my biggest beef with OpenClaw is the same beef I have with many AI systems: it’s a giant black box. In other words, even its own developers don’t always understand why it’s doing what it’s doing. This is a common problem in the AI development world. When you start to give AI the ability to go beyond simple things like writing text and you want it to start reasoning and making decisions, many of the systems are capable of that, but it’s not always clear how they’re capable of it and what precise mechanisms are leading to those decisions. Where was the root of the problem? Was the LLM it’s using putting out bad output? Was the problem a plugin that was being used? Was it a context problem? Does the prompt need adjusted? Did a human feed it bad information at some point in the workflow? These answers are frequently quite blurry when dealing with many AI reasoning models, and definitely with OpenClaw.
In property management, explainability beats cleverness almost every time. I don’t need my AI to be creative with its accounting tasks. I just need it to do precisely what I want it to do every time. Creativity is risky behavior, especially when it’s occurring behind an opaque wall.
OpenClaw Just Adds Unnecessary Complexity
People sometimes think that open source software is simpler by default because it doesn’t have a giant corporation looking to add a bunch of bells and whistles to it to sell it to the largest possible demographic, but that’s not always the case.
In reality, when a system like this gets the attention of software developers, the complexity starts to stack on pretty quickly as all of them want to do different things with it. Imagine a giant collection of hundreds of thousands of the nerdiest people on earth suddenly realizing that they have a new toy to play with. Very exciting for them, very risky for you.
By contrast, more “boring” and tried and true systems can do pretty much everything we want to do in the world of property management without the need for any of this additional complexity and risk. All we really need is:
A constantly running server to make sure the system is working around the clock
A task queue to intake work it needs to do for you
Webhooks to tell it what to do and when
Some form of database (which could be as simple as your property management accounting software if it has an open API)
Small scripts written for each workflow
A highly secure environment
Tightly controlled API connections
None of that needs anything even remotely like OpenClaw. Instead, you just need a simple system like a VPS (virtual private server) and an AI coding platform to write the specific code to that VPS for what you want it to do. All of that open-ended reasoning capability is overkill, and frankly, frightening.
Cost Discipline Matters
OpenClaw may seem cheap because it’s free to install, but the up-front cost isn’t the real cost. The real cost comes from how it behaves. Because OpenClaw runs autonomously once you really get it going, it can be working around the clock doing all sorts of things and racking up usage costs. I asked my OpenClaw instance to research the best way to build a workflow, and it just kept thinking…and thinking…and thinking…and thinking. It got stuck in a loop trying to find the best way to do something that a simple LLM could have solved in mere seconds. After a couple of hours of pointless “reasoning,” I finally just told it to stop and I asked Claude. Had I not stopped it, it may have just kept confusing itself in this endless loop forever while burning through tokens (money).
By contrast, a custom-built stack that you develop yourself from the ground up does only what you specifically want, when you specifically want it, and the cost is almost non-existent. People are running around scooping up Mac Minis at a cost of $600 a pop while I just got a Linode VPS at a cost of $25/mo and I don’t have to worry about keeping a local computer running 24/7. It can do anything that any sane PM would ever want their AI automation platform to do.
Custom Means Boundaries and Safety
The best AI systems in our industry aren’t going to be open-ended and autonomous; they’re going to be rules-focused and bounded. We don’t need some robot constantly roaming the internet for us. If you want that for your personal assistant, great, do that, but don’t ever put it on any computer that has access to any of your PM data.
What you want to do is custom build workflow automations for each task you want the AI to complete for you, and then just push those automations to the VPS where they will do the work for you 24/7/365, but only when specifically called to do so by your systems. And if something can have irreversible negative effects (for example, reporting on someone’s credit, or filing an eviction against them), make sure the AI can’t act without first getting an approval from a knowledgeable human.
I would contrast these two models like this:
OpenClaw - “Give the agent access to communications, files, and tools so that it can manage all of the work itself.”
Custom VPS Build - “Create separate services for maintenance triage, owner update drafting, lease drafting, and inbox management, each with clearly scoped permissions.”
The latter is far safer, easier to test, easier to fix and replace, easier for your team to understand, and easier to improve incrementally.
Architecture Independence Matters
How many PMs do you know who are still using the same PM accounting software that they hate, but they’ve been using it for years and it has all of their data, so they feel trapped? Well, welcome to people a few years from now who have trapped themselves into OpenClaw’s environment and now it’s just too much work to get out of it.
OpenClaw may seem less like a software architecture prison because it’s open source, but it’s really not. It’s still someone else’s architecture. You are still dependent upon that foundation.
By contrast, when you custom build, you own not only the data, but the entire platform. Imagine if OpenClaw changes direction a couple of years down the line and the platform no longer works well for you. Sure, it’s open source, which means you could theoretically contribute your own code to the project and make it do what you want it to do, but adding your own code to an open source project is a far bigger lift than simply vibe coding some simple API integrations. If you aren’t a software developer, you’re going to have a very hard time bending another platform, even an open source one, to your will.
The bottom line is that property management tech stacks are already fragmented more than enough. The last thing that we need is another layer of software between us and what we actually want done that becomes difficult to unwind later. Especially when there’s just no valid reason for it. At least the PM accounting software platforms are a basic necessity. You can’t have a PM company without trust accounting, owner and tenant portals, etc. But OpenClaw is just not necessary, and creates many new problems.
An Ode to Linode
I’ve mentioned the term “VPS” several times already in this article, so let’s talk about what that really is. A VPS (virtual private server) is just a computer that lives in the cloud (or more accurately, a whole bunch of computers). You essentially lease that server at a monthly cost and can do whatever you want with it. It’s not magical, but it’s useful because it’s so simple.
It offers predictable and cheap pricing (I pay $25/mo which includes automated backups and that’s for overkill server power), more than enough power for anything you’ll ever need to do, it’s secure, it’s redundant, and it’s running 24/7/365. And you can just keep upgrading with a click of a mouse if you ever need to (and you probably won’t ever need to). I use a provider called Linode, and they offer everything from bare bones $5/mo shared CPU servers, all the way up to dedicated GPU options for those doing super-intense AI processing. You will never need the latter, no matter how complex and advanced you want your PM AI to be.
And with another click of the mouse, you can just add another server if you want each AI agent to live on its own machine. I currently run two Linodes, and I’ll probably have half a dozen before it’s all said and done. I like keeping different roles separated on different servers so it’s easy to track what’s happening, and I can setup a main server to coordinate between all of them.
Keep it boring. Boring is safe. Boring is predictable. Boring just works. And Linode is plenty boring in the best possible way.
OpenClaw Still Has a Place
For everyone out there who has just been absolutely wowed by what they’ve seen online from AI influencers and their OpenClaw instances, I’m not telling you that you can’t use OpenClaw for anything or that it doesn’t have a valid purpose. I’m just telling you that its place isn’t in your property management company’s operations.
OpenClaw has real potential for techie people who want to tinker with powerful autonomous AI and put it to general personal use. As a personal assistant, there probably isn’t a better option out there yet. But you still need to be really careful with it. You need to make sure that you put it on a completely separate computer than your own, make sure it has no access to any of your personal data that you don’t specifically give it, and don’t ever give it access to any of your business data or systems. It should have a dedicated email account, web browser, etc. that you don’t use for any other purpose. You don’t want it accessing your own email, because once you give it that power, the possibilities for havoc are limitless. It can have its own email account that it uses for doing work for you.
When it comes to your PM business, keep it separate, and keep it boring.
Final Thoughts
I’m not saying that OpenClaw isn’t interesting or impressive. It certainly is. But impressive does not necessarily equal appropriate. We always need to select the best tool for the job, and not just pick up a shiny new hammer because it’s shiny and easily available. The smartest AI strategy for PMs is probably not “a really powerful AI agent that can do just about anything.” In property management, the future likely instead belongs to AI systems that are less theatrical and more tried and true. And yes, boring.
Open to Work
Are you an experienced PM industry employee looking for work? Or are you a PM company or vendor seeking the best talent? Send me your info and I’ll feature it here!
Each week we also feature a Remote Team Member available through VPM Solutions. See this week’s below! Click on the image to get more info:
We always have ad slots available for industry vendors. Sign up for an ad or ad package here. Pricing and ad details can also be found at that link. Questions about advertising? Email abio@propertymanagerassist.com or book a time here.
Debate Me
Disagree with my take here? Have a different perspective? There’s nothing I love more than a good debate or even just an intelligent conversation. If you’d like to jump on a podcast recording with me to discuss this topic, please let me know!
The views expressed in this publication are the views of the author only and not any advertisers, sponsors, partners, affiliates, or organizations that the author may be a member of.
PMAssist Industry Insights is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
