A software integration is a standing agreement between two programs about what data moves between them, in which direction, and when — handled automatically instead of by a person retyping it. When a proptech vendor says a tool “integrates with Yardi” or “connects to your CRM,” that is what they mean: the two systems pass information to each other without someone copying it by hand. For a 4–20 person commercial real estate firm running on Outlook, Excel, a proptech subscription, and a folder of lease PDFs, the word usually shows up in a sales pitch or a price quote — so this guide explains what you are actually being sold, the four honest ways two tools get connected, what it costs to keep running, and when the right answer is not to integrate at all.
What a software integration actually is
Picture two apps and a translator standing between them. The first app speaks its own language, the second speaks another, and the translator carries agreed-upon facts across the gap — a new tenant, a signed lease, a paid invoice — so nobody retypes anything.
That picture holds up better than the technical definitions elsewhere. An integration is not a feature you switch on once. It is an ongoing arrangement: a defined set of data, a direction of travel, a trigger for when it moves, and something in the middle doing the carrying.
Three questions define any integration, and you can ask all three in plain language:
- What data moves? A contact, a rent roll line, a document, a dollar amount.
- Which direction? One-way (your CRM feeds your accounting tool) or two-way (both stay in sync).
- When does it move? The instant something changes, on a nightly schedule, or when you press a button.
If a vendor cannot answer those three questions crisply, the integration they are describing is vaguer than it sounds. A firm that runs on the kind of Excel-and-Outlook stack most small shops do — the setup we break down in our look at the anatomy of a small firm’s tech stack — has more of these connection points than it realizes, most of them still handled by a person.
Why the word shows up everywhere in proptech
“Integration” is a selling word because it promises to erase the most hated task in a small firm: entering the same information twice. Every proptech vendor knows that double entry is where your day leaks away, so every product page lists the tools it “connects” to.
The problem is that “integrates with” covers a huge range. It can mean a deep, maintained, two-way sync — or it can mean the vendor will let you export a spreadsheet you then import somewhere else by hand. Both get the same word on the marketing page.
That gap is why a plain-English definition matters before you buy anything. When a CoStar, Yardi, AppFolio, Buildout, or Dealpath listing says “integrations,” your job is to make the salesperson tell you exactly what data moves, which direction, and how often. The answer determines whether you are buying real automation or a fancier export button.
This is the same discipline that runs underneath every buy-versus-build decision, which our buy-vs-build playbook for off-the-shelf proptech covers end to end. Integration is one of the places where a demo looks effortless and the real work stays hidden.
The four ways a small firm connects two tools
There are only four honest ways two pieces of software end up sharing data at a firm your size. Knowing which one a vendor is proposing tells you the cost, the fragility, and who has to babysit it.
| Method | What it is | Cost shape | Who maintains it | Best for |
|---|---|---|---|---|
| Built-in (native) | The vendor already built a connection to a named partner tool | Usually included or a small add-on in the subscription | The vendor | Common tool pairs (a CRM and a mainstream accounting package) |
| No-code connector | A service like Zapier or Make wires two apps together with rules you set, no programming | Low monthly subscription, often under a few hundred dollars | You (plus the connector service) | Light, non-sensitive glue — a form to a spreadsheet, an email alert |
| Export / import | One tool spits out a CSV or PDF; a person loads it into the other | “Free” but paid in your hours | A person, every time | Low-frequency transfers where a human check is welcome |
| Custom pipeline | Software built specifically to move your data between systems | A one-time build, typically in the $25,000–$150,000 custom-automation range, plus modest monthly run costs | Whoever built it (you keep paying to keep it alive) | Workflows no product connects, or ones that cross systems no vendor bridges |
Most small firms overestimate how often they need the fourth option. A built-in connection or a no-code connector handles a large share of everyday needs, and the export-import route — unglamorous as it is — is often the correct answer for something you do twice a month.
The custom pipeline earns its place only when a workflow is specific to how your firm makes money and no product fits it. That is a real scenario in commercial real estate, where deal structures and reporting formats vary firm to firm — but it is the exception, not the default.
What an integration really costs
The sticker cost of an integration is almost never the real cost. The real cost is that integrations break, and someone has to fix them.
Software vendors change how their systems talk to the outside world on their own schedule. When they do, a connection that worked yesterday can quietly stop moving data today — and the failure is often silent, so you find out when a number is wrong two weeks later. Every integration you own is a small ongoing maintenance obligation, not a one-time purchase.
This is where the four methods diverge sharply:
- A built-in connection shifts that maintenance to the vendor. That is the main reason to prefer it — you are paying them to keep the pipe working.
- A no-code connector puts light maintenance on you: when something changes, you adjust the rules, which is usually manageable but not zero.
- A custom pipeline is the expensive one to keep alive. You own the upkeep, and a build that no one is paid to maintain becomes technical debt that fails at the worst moment.
Price any integration across twelve months, with your own time counted, not as a one-time line item. A cheap connector that eats two hours a month is not cheap, and a custom build with no maintenance plan is a liability wearing the costume of an asset — a trade-off our guide to how AI automation projects work, from discovery to deployment walks through against a real project timeline.
The confidential-data question you have to ask
Every integration moves data through something, and in commercial real estate that data is confidential — deal terms, rent rolls, tenant financials, signed leases, often under NDA. The moment two systems share it automatically, a third party frequently sits in the middle carrying it.
Before you turn on any connection that touches deal data, get three answers in writing:
- Where does the data go, and who can see it? A connector or pipeline usually stores or passes your information on someone else’s servers. You want to know whose.
- Is my data used to train a shared model? For any connection that involves an AI feature, you want a clear no. Your confidential deal data should not become training material for a product other firms use.
- How do I turn it off and delete everything? A connection you cannot cleanly sever, with data you cannot fully export and delete, is a connection you do not control.
A no-code connector moving public marketing data is low-stakes. The same connector moving a tenant’s financials is a different decision entirely. Match the scrutiny to the sensitivity of what flows through the pipe, and treat a vague answer on data terms as a reason to keep the transfer manual.
None of these are technical questions. They are business and contract questions an owner can ask directly — the same posture that lets a small firm buy AI safely without an IT department, which is part of what deciding whether to bring in an outside AI partner is really about.
When not to integrate
The honest answer for a lot of small-firm workflows is: do not integrate this. An integration is worth building when the same data crosses between two systems often enough that the automation saves more time than it costs to maintain. Below that line, simpler wins.
Two alternatives beat a fragile integration more often than vendors admit:
- A clean manual export on a cadence. If you move data between two tools twice a month, a fifteen-minute export-and-import — with a human eye on the numbers — can be cheaper and safer than a connection that breaks silently and needs upkeep.
- A prompt over a tool you already own. For pulling structured facts out of a lease PDF or drafting a summary, a well-written prompt in ChatGPT, Claude, or Gemini often does the job without any integration at all, using software you already pay for.
The test is volume and stakes. High-frequency, high-value, error-prone-by-hand workflows justify a real integration; occasional or low-stakes transfers usually do not. Reaching for a custom build there is how small firms end up paying $25,000-plus to automate something that happened four times a year.
Start from the workflow that hurts, not from the tool a salesperson is holding. That ordering — problem first, connection second — is the whole of the small-firm AI operating philosophy that lets a lean shop move faster than a firm ten times its size.
Frequently asked questions
What is a software integration in simple terms?
A software integration is an automatic arrangement for two programs to share specific data so a person does not have to retype it. It has three defining parts: what data moves, in which direction, and when. When a proptech tool says it “integrates with” your accounting software, it means the two exchange information — a new tenant, a paid invoice — without manual copying. The depth varies widely, from a real-time two-way sync to little more than an export button, so always ask a vendor exactly what data moves and how often.
What is an API, and do I need to understand it?
An API is the doorway one piece of software opens so other software can request or send data in a controlled way. You do not need to understand how it works, only that it is the mechanism most integrations use to talk. The practical point is that APIs change on the vendor’s schedule, and when they do, an integration built on one can break. That is why maintenance — someone keeping the connection working — is part of the real cost, not an optional extra.
How much does a software integration cost for a small firm?
It depends on which of the four connection methods you use. A built-in integration is usually included in a subscription or a small add-on. A no-code connector like Zapier or Make is typically a low monthly fee, often under a few hundred dollars. Export-import is “free” but paid in your hours. A custom-built pipeline generally falls in the $25,000 to $150,000 range for commercial automation, plus ongoing run costs. Price any option across twelve months with your own time included, because the sticker is rarely the full bill.
What is the difference between a native integration and a connector like Zapier?
A native integration is one the vendor built and maintains between their product and a named partner tool, so upkeep is their job. A connector like Zapier or Make is a separate service you configure to wire two apps together with rules, and the light maintenance is yours. Native is the safer default when it exists for the tools you use, because you are paying the vendor to keep the pipe working. A connector is the flexible option for pairs no vendor connects natively, best kept to lower-stakes data.
Is a no-code tool like Zapier safe for confidential deal data?
Treat it with caution. A no-code connector routes your data through a third-party service, so for confidential deal information you need to know where the data is stored, who can access it, and whether it trains any shared model. For public marketing data or internal alerts, a connector is low-stakes and fine. For rent rolls, tenant financials, or NDA-covered deal terms, get the data terms in writing first, and default to a manual transfer if the answers are vague.
Why do software integrations break?
Integrations break because the systems on either end change independently. When a vendor updates how its software shares data, a connection built on the old method can stop working — and the failure is often silent, so you notice only when a figure is wrong. This is why every integration is an ongoing obligation, not a one-time setup. A built-in integration puts that maintenance on the vendor; a custom pipeline puts it on you, which is why an unmaintained custom build is a common source of expensive surprises.
Do I need a developer to set up an integration?
Not for most of what a small firm needs. Built-in integrations are turned on inside the software with no code, and no-code connectors are designed for non-technical users to configure. You only need a developer or an outside partner for a custom pipeline specific to your firm that no product offers. The judgment calls that protect you — pricing the true cost, demanding data terms in writing — are business decisions, not technical ones, and a few hours of fluency training helps far more than a hire.
When should a CRE firm build a custom integration instead of buying a tool?
Build a custom integration only when a workflow is specific to how your firm operates, no product connects the systems involved, and the data crosses often enough that automation saves more than it costs to maintain. Most 4–20 person firms should default to a built-in connection or a no-code connector and earn the custom-build case with real volume data. A custom pipeline is justified for high-frequency, high-value work that is error-prone by hand — not for a transfer that happens a few times a year.
Can I just use ChatGPT or Claude instead of integrating my tools?
Often, for the right tasks. For pulling structured information out of a lease PDF or drafting a market summary, a well-written prompt in ChatGPT, Claude, or Gemini can do the job at low volume using software you already own, no integration required. The limits: general AI tools do not connect to your systems or keep anything in sync, so you still move the result by hand, and you must verify every figure before it drives a decision. Use them to prove whether a real integration is worth building at all.
Where to start
Before you say yes to any integration a vendor is selling, it helps to know which of your workflows actually deserve one — and in what order. Most firms shop for connections tool-first, when the honest sequence is problem-first: name the double-entry task that costs you the most, then decide whether the fix is a built-in feature, a connector, a manual export, or a custom build.
A free AI-readiness assessment produces exactly that map. It is a working session that inventories where your data gets retyped, flags the connections that would pay for themselves, and hands you a ranked plan with real cost ranges before you sit across from a salesperson. Book a free AI-readiness assessment if you want that picture first. If the right answer for your firm is to keep something manual, the assessment will tell you so — and you will still leave with the plan.
Dirk Jan van Veen, PhD
