Veldi / ai for property management
Written from inside one. We run a Utah property management company on AppFolio, and we built our own automation stack because nothing on the market did what we needed. This page is what we learned about where AI actually returns hours in this business, and where it quietly wastes your money.
Most PM companies already own good software. AppFolio, Buildium, Rentvine and their peers handle the ledger, the lease, the owner statement, and the compliance trail well. Ask a property manager where their week goes and almost none of it is the software. It's the coordination that happens between the software and the world.
A tenant texts about a leak at 9pm. Someone has to read it, decide whether it's an emergency, figure out which plumber is both available and approved for that owner, check whether the last visit was under warranty, get owner approval if it's over the threshold, dispatch, confirm, and then log all of it back into the platform. The platform records the work order. It does not do a single one of those decisions.
Multiply that by every maintenance request, every leasing inquiry, every renewal decision, every owner who wants to know what's happening at their property, and you have the actual job. That coordination layer is where the hours go, and it's the layer AI is genuinely good at now.
The useful mental model: your PM platform is the system of record. What you're missing is a system of action sitting on top of it.
Ordered roughly by hours returned per dollar spent, based on what actually moved the needle in our own operation and our clients'.
Incoming request comes in by text, email, or portal. The system classifies the trade, judges urgency against your rules, checks warranty and recent history at that unit, picks the vendor who actually handles that trade for that owner, drafts the dispatch, and logs it back to the work order.
This is the single biggest win in most PM companies. It's high volume, it happens after hours, and the decision logic almost always lives in one coordinator's head, which makes it a staffing risk as well as a time cost.
Speed to lead decides who rents the unit. An automated first response that answers the actual question asked, screens against your criteria, and offers real showing times converts dramatically better than a form autoresponder.
The measurable outcome here isn't hours saved, it's days off market. On a $1,800 unit, cutting five vacant days is worth about $300 per turn, every turn.
Owners don't churn because of one bad month. They churn because they feel uninformed. Automated monthly narrative summaries, plus a proactive note when something notable happens at their property, changes that relationship without adding a person.
This is also the cheapest retention lever in the business. Every owner you keep is a full acquisition cost you don't spend.
Ninety days out, the system flags the expiring lease, pulls comparable market rent, weighs the tenant's payment and maintenance history, recommends a renewal rate, and drafts the offer for a human to approve and send.
Most companies leave real money on the table here purely because nobody had time to research each unit. It's a research problem, which is exactly what AI is for.
Your platform already emails you plenty of reports. Almost nobody reads all of them. A monitor that parses them on arrival and surfaces only the exceptions (work orders aging past your SLA, units vacant too long, expiring insurance, delinquency crossing a threshold) turns passive reporting into an actual alert system.
Low build cost, and it catches the expensive things while they're still cheap.
Onboarding is a long, mostly clerical checklist: gather details, pull the tax record, set up the property and units, build the listing, produce the agreement, schedule the initial inspection. Much of it is assembly from information you already have.
Automating it matters most when you're growing, because onboarding is exactly the thing that breaks when you sign three owners in a month instead of one.
Selling automation is easy. Knowing what to leave alone is the part that keeps you out of trouble, and in property management the downside is regulatory, not just operational.
Applicant screening decisions, and any language that could be read as steering, need to stay under documented human control with consistent written criteria. Use AI to gather and organize application materials. Do not use it to decide who gets the unit, and never let it improvise the wording of a rejection.
Three-day notices, lease violations, eviction filings, and security deposit dispositions all carry statutory deadlines and format requirements that vary by state. Draft with AI if you like. Send only after a human who knows your state's rules has read it.
Owner distributions, refunds, and vendor payments should stay behind human approval. Automate the preparation and the reconciliation, not the release.
Nobody should be rebuilding general ledger and trust accounting to get better automation. Your platform is the record. Build the action layer around it and write results back into it. That decision alone avoids most of the expensive mistakes we see.
Good rule: automate the work, keep the human on the judgment that carries legal or financial consequence. That's not a limitation, it's what makes the rest of it safe to run unattended.
Buy where the category is mature and your process is standard. Leasing chatbots, listing syndication, tenant screening, and renters insurance tracking all have credible off-the-shelf products. There is no advantage in building your own.
Build where the workflow spans several systems that no vendor connects, or where the process itself is how you win. Your vendor dispatch logic, your owner communication style, and your onboarding sequence are all specific to you, and they're exactly what a generic product flattens.
The failure mode we see most: a company buys six overlapping tools, each solving about fifteen percent of the problem, none of which talk to each other. Now you're paying six subscriptions and the coordination burden went up, because someone has to keep them all in sync. One connected system built around how you actually work beats six that don't.
This is the entire reason we productized the assessment. Two weeks, a flat fee, and you get a ranked list of what's worth automating in your specific operation with a working prototype of the top item, running on your data. If the answer is "buy these two products and don't build anything," we'll tell you that, and you'll still have the roadmap.
No, and it shouldn't try. AppFolio, Buildium, Rentvine and similar platforms are your system of record for ledgers, leases, and compliance. Replacing them is enormous risk for no real gain. Automate the layer above them: triage, follow-up, coordination, and reporting, then write results back into the platform.
Maintenance work order triage and dispatch, in almost every case. High volume, happens at all hours, has clear decision structure, and delay costs real money in both tenant satisfaction and property damage. It's also usually sitting in one coordinator's memory rather than in any system, which makes it a staffing risk too.
For a 150 to 600 unit portfolio, 10 to 25 hours a week across the team from the first two or three automations is a realistic target. It depends much more on how much of your process currently lives in email and text than on your unit count.
It can be, with deliberate design. Keep the system of record in your PM platform, never send full PII to a consumer chatbot, use business-tier API agreements with no training on your data, log every automated action, and keep a human approval step on anything touching money or a legal notice. We design to these rules by default.
Yes. We run our own company on AppFolio. In practice the integration surface is a mix of report exports, email-delivered reports, and the browser itself, and a well-built automation layer uses all three. The same approach applies to Buildium, Rentvine, and Propertyware.
Below roughly 100 units it's usually premature, with one exception: if you're growing fast and about to hire your first coordinator, automating first is often cheaper and more durable than hiring. That's a real conversation to have, and we'll tell you honestly which side of the line you're on.
Two weeks, flat fee, a ranked list of what's worth automating in your operation, and a working prototype of the top one.