AI-First or AI-Ready? The Property Management Process Debate
One approach says to redesign operations from the ground up. Another says to organize today’s workflows so AI can gradually take over. Both may be right.
Back in April I read two articles that Peter Lohmann had linked from his newsletter, both talking about how property managers should be thinking about AI in terms of their processes. One article argues that property management systems need to be redesigned from the ground up with AI-first principles in mind. The other article argues that you should structure today’s already-existing processes so that AI can just take over the parts that AI is best at without a full redesign. I’ve been meaning to write about this because I think it’s such an interesting debate, but other topics kept coming up. Finally, today, we’ll cover this in-depth. But I strongly recommend reading both articles in full.
The first article, written by AI vibe-coding master Wolfgang Crosky, advocates for going back to first principles and designing your processes from scratch again with an AI-first mindset.
PMAssist Industry Insights is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
The second article is written by Rob Lowry, Founder of an AI consulting firm called LaunchEngine. Rob argues that property managers are basically just too busy and in the weeds to redesign their entire processes, and a more measured and piecemeal approach makes more sense.
Ultimately, both agree that AI is the future. We’re eventually going to end up in a place where AI-first is king. But how we get there is definitely an interesting discussion, so let’s dive in.
Position 1: Start from First Principles
Wolfgang and I are cut from the same cloth on this topic. We’re both practically addicted to vibe-coding, and we have made enormous changes in our businesses centered around AI adoption. So you would think I would come down firmly on his side in this debate. But I think it’s a bit more nuanced.
Wolf argues in his article that this isn’t just about automating checklists. A workflow shouldn’t be just a list of steps to follow; it should explain the intent, it should include guardrails, it should take into account the emotional experience of all involved, it should meet compliance requirements, and it should specifically delegate only certain items where judgment is necessary to a human in the loop (HITL in nerdspeak).
Wolf’s argument is compelling, especially to someone like me who is quite technologically inclined. The reality is that existing processes in our business tend to be built around deficiencies in software and humans. If you look through the steps in your processes, you can probably find dozens, if not hundreds, of examples of steps that are only there because you need to find a “workaround” in your software because it doesn’t work the way you want it to. And of course every process is filled with checkpoints to make sure that the imperfect apes that we call humans don’t screw things up. These things scream out for the advances of AI to solve these problems.
But on top of that, if we want to extract maximum efficiency from our businesses, we need to allow the AI to do its thing as a semi-autonomous agent instead of it just running automations in the background based on human task completions. The problem with that is that AI agents need more to go on than just a list of tasks. AI works best when it has context, when it knows what the end goal is, and when it understands the various pitfalls that may be involved. And yes, I know that some of you blanch at me using a word like “understands” when it comes to AI, because you think that AI is just a glorified autocomplete system, but I would challenge you to go ask complex philosophical questions to Claude and engage it in a conversation. I would argue that Claude “understands” just the same as you and I. We’re just “wired” on meat instead of on circuit boards and software. The underlying mechanisms of understanding may be different, but the fact that understanding is taking place in some form really can’t be denied at this point. Any argument you can make that AI doesn’t understand what it’s saying can be repackaged and used as an argument that humans don’t truly understand anything either and are just engaging in deterministic outcomes that are too complex for us to understand. The bottom line is that whether you believe that AI is truly intelligent or not, the outcomes that AI produces are consistent with something that understands, whether it truly does or not.1
So, with this framing that AI needs to understand the bigger picture rather than just having a list of tasks to complete, it’s pretty easy to see how Wolf jumps to the idea that entire processes need to be rebuilt with AI in mind from the foundation up. We aren’t just trying to make tasks more efficient, we’re trying to get rid of the tasks altogether and allow the AI to handle things autonomously behind the scenes within a framework of guardrails and context. Basically, Wolf is telling you that most of us are asking the wrong questions. We’re asking things like “how can we automate this” or “how can we make this task more efficient,” when what we really should be asking is “should this process even exist in this old form at all?”
You’re staying on your current platform because switching feels risky. Meanwhile, it just keeps getting worse.
Rentvine makes the switch easy and the upgrade obvious. We handle 80% of your migration, complete your first reconciliation, and get you live in an average of 45 days.
Here’s what changes the day you go live:
Month-end close drops from a full day to 1 to 3 hours
Support responds in 35 minutes on average, with unlimited training for life
AI is built in from day one across vendor bills, maintenance coordination, and leasing
Your portfolio finally has room to grow, no workarounds, no side spreadsheets.
Recently ranked the #1 property management software in the 2026 VPI Category Report, the risk isn’t switching…it’s staying another year.
In his article, Rob takes a different approach to all of this. Not that he sees the ultimate destination any differently; he just sees the path to getting there quite a bit differently.
His thesis is pretty sound. The idea is that the average property manager is so overloaded with admin work and putting out fires all day that they simply don’t have the capacity to just pause operations and rebuild everything from the ground up to accommodate AI.
Instead, Rob argues that we need to take this one step at a time. First, workflows need to be tweaked so that they have clearly visible status markers, clearly documented handoff points, automated execution of most steps, and a clean flow of data for auditing purposes. Basically, the goal here isn’t to replace every layer of the workflow all at once; the goal is to take a deliberate approach to improving the business piece by piece by implementing AI where it makes sense.
This is clearly framed as a more pragmatic approach that fits into the operational reality of the typical property management company more easily. Instead of this being a big transformational shift overnight, it’s a slow moving towards the ultimate goal in a way that a busy property manager is more easily able to handle. Think of this as AI just being a new employee that needs to be onboarded and slowly trained on everything they need to know and do. The new employee doesn’t immediately hit the ground running at full speed on day one on the job. Realistically, it takes months for an employee to truly feel comfortable in a new role, and years to be actually a master at that role.
Points of Agreement
While both authors are taking different paths, they’re both ultimately looking for the same goal at the end, and that means they end up agreeing on quite a bit.
First, both argue that AI isn’t going to be a magic fix to operations that are already a mess. I don’t do much consulting work nowadays, but I did it nonstop for several years, and that allowed me to see under the hoods of a lot of PM companies across the country. And not to be insulting, but a lot of you are running a complete mess. Processes are all in your head, policies are “case-by-case,” every day at the office is barely-controlled chaos, etc. In such an environment, adding a little bit of AI to the mix isn’t going to solve anything. Instead, all that adding AI will do is further expose just how much of a house of cards you’ve constructed there.
Second, while both authors are looking at documentation a little bit differently, both are still on the same page that documentation of some sort is a necessity. You can’t get anywhere with AI if you don’t know your systems well enough to have them documented. That documentation could be what Wolf describes with intent, judgment, etc. all baked in, or it can be documentation that is just triggers, tasks, and stages like Rob discusses, but either way it’s documentation. You can’t build reliable AI into your company if you don’t first have documented what actually needs to happen.
Both authors also agree that this isn’t about fully eliminating humans. I don’t know where they fall on this debate, but I’m certainly someone who believes that it will eventually be possible to have humans doing almost nothing that they don’t simply want to do. The AI will eventually reach the point that it is not only as good as an extremely experienced human, I think it will very quickly reach the point of being superhuman and outperforming humans at even the most complex of decision-making and judgment. But we certainly aren’t there yet. Having a human involved at various stages of your processes is still very much a requirement.
Finally, both authors are in agreement that AI isn’t simply a glorified chatbot or writing assistant. AI is something that is going to actually participate in PM workflows at some level, whether that’s as an autonomous agent or just as a more efficient task completer. But if you’re still believing that AI is worth nothing more than writing some blogs for you and answering basic questions in a chat on your website, then the authors and I would all agree that you are wildly out of step with reality at this point. AI passed that point long ago.
Points of Divergence
This is where things are truly interesting. Both authors make great arguments, but those arguments differ in very important aspects and conclusions. Let’s look at these disagreements individually.
First, where do you start? Wolf is telling you that you need to start from the ground up and redesign your process with the idea of AI firmly embedded in the entire process from the beginning. This isn’t an exercise in just adding AI to the existing mix, it’s a complete restructuring of the whole process with AI in mind from the beginning. On the other hand, Rob is saying that you should start where you are and keep it more simple. Make the necessary changes to the existing process to fit in AI, but you don’t need to radically redesign everything.
I think this mostly comes down to a philosophical question. Is this design-first, or implementation-first? In a design-first philosophy, you look at what the ultimate end result of the process should be and work from the foundation up to reach that goal, building it all from scratch. In an implementation-first philosophy, you are focused primarily on pragmatism and making sure the business is not disrupted while you make changes.
But I think this also comes down to a question of risk tolerance. Someone who is very risk-averse is going to be quite hesitate to engage in a full redesign of their processes. That requires a very deep organization change while Rob’s methodology is closer to just tweaking what already exists. To someone who is risk-averse, the latter is far more comforting.
Finally, I think the disagreement comes down to just how the two authors would define “readiness.” For Wolf, readiness is about making sure that your workflows are build around purpose, trust, emotional understanding, and continuous improvement. For Rob, readiness is a little more structural and less philosophical. It’s about things like transparent workflows, automated execution of tasks, clearly defined decision points, etc. I would say that Rob’s approach is less holistic, but more realistic. At least for the average property manager.
Who’s Right?
I hate to disappoint, but I’m not going to just declare one author a winner here. Because the reality is, different companies would do well to take different approaches to this problem.
I don’t see Wolf as being the average property manager. Even before AI arrived on the scene in a big way, Wolf’s company was already highly automated and thoroughly systematized. I didn’t know what a Zap was until I heard Wolf talking about Zapier years ago. He was really the person who introduced all of this automation craze to the PM space. So he was already light years ahead of virtually everyone else when ChatGPT first dropped and AI entered the scene for ordinary people.
The reality for the average small and mid-sized property management business (call it everything under 1,000 doors) is a lot different. Most of them still don’t even have any Zaps or any other kind of automation that isn’t directly built in to their primary PM accounting system. Hell, most of them don’t even have checklists for their core processes! There are tens of thousands of property management companies across this country. Only a few thousand of them are NARPM members and regularly engaged in these discussions about the latest and greatest ideas and technology in property management. The other tens of thousands are just trying to get through the day without anything burning down. For these companies, Wolf’s approach simply isn’t possible. They literally don’t have the time in the day to sit down and think from first principles on process design. If they did, properties wouldn’t get leased and maintenance wouldn’t get done.
So, if I’m to play the role of the judges in this metaphorical boxing match, this is a split decision. There is no knockout. What works best for one company is different than what will work best for another. In my own business, I’m even doing it differently for different processes. Some processes I’ve taken Wolf’s approach, and some I’ve taken Rob’s. You might end up doing the same.
A Framework for the Undecided
If you’re reading this and you’re on the fence, not sure which approach makes more sense for your business, then I think we can put together a framework that blends the best of both worlds.
Document first. Before you do anything, if you don’t already have how you do things in your process documented in some fashion, then you need to get this done first. You can’t make any rational decisions at all on how to proceed if your process isn’t taken out of your head and committed to paper or bits on your computer.
Put that documentation into a structure. Divide everything up into statuses, tasks, handoffs, and decision points.
Determine HITL points. You will need to identify the areas where you simply don’t trust automation or AI to handle things and you want a human in the loop. Flag those and keep them assigned to a human. If you end up with too many of these, though, you might need to reexamine any irrational fears you might have about AI. It’s rational to want a human to read marketing descriptions prior to publishing to avoid fair housing violations. It’s not rational to want a human to personally review every late rent email that goes out before it’s sent.
Automate. This isn’t the same as autonomous AI. This is just deterministic automation. If this happens, do that automatically. This what you do with the task items from step 2 above that don’t require a HITL.
Redesign for customer experience. After you’ve gone through the above steps, now you have freed up some time and can devote more efforts to broader redesign. You can start to think about the things that Wolf emphasized like the customer experience and how something impacts them emotionally. You weren’t able to devote resources to this before because you had too much work to do. Now that you’ve passed that off to the automation, it’s time to dig deeper.
Introduce semi-autonomous AI agents. I say “semi-autonomous” rather than straight autonomous because this business is just too litigious for 100% autonomy for AI. Unless regulators give us some sort of safe harbor where AI is involved and promises not to violate us for AI mistakes (a highly unlikely thing for regulators to do ), then we need to keep some pretty strict guardrails on AI agents and be sure to monitor them closely. This is not the time to throw OpenClaw at everything and just let it do its thing. I promise you, that will be a disaster. Slowly start to delegate things to the AI agents like preparing email responses, but not sending them until a human reviews. Slowly release the reins on the agents over time as you grow more and more confident that it isn’t going to get you in trouble. And there are some things that you’ll never give the agent full autonomy over.
Continuous improvement. This isn’t something that has a final end state. You want to continue to monitor performance, spot-check autonomous interactions, and verify that things are happening the way you want them to happen. As you discover gaps or areas that can be improved, make those tweaks. You will have fewer and fewer to make over time, but it will never completely end.
AI Will Reward the Best Operators
You can’t just throw AI at your systems and expect it to make your company better overnight. If you do that, AI is simply going to amplify any weaknesses that already exist in your business. Instead, you need to take a structured, deliberate approach to this. That approach could be closer to Wolf’s, or to Rob’s, or to my blended approach. But it needs to be a structured approach.
As I keep saying in this publication, the future of AI in property management is not about replacing property managers. There will always be a need for the work that we do, and humans will always be a part of that. But AI will be about separating the wheat and the chaff and allowing the best operators to rise to the top of the pack. Property managers who steadfastly refuse to take part at all in the AI game will end up going out of business, and property managers of varying levels of AI inclusion will find different business models for different customer segments to make it all work together. Find the approach to it that works best for your business and get started on it. Now is the time.
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.
“Understanding” is a different concept than consciousness, to be clear. I’m not arguing that AI is conscious (although I think that AI will eventually reach that point in the future).
