A property manager in Hamilton called me three weeks ago. Her property-management software vendor had just announced an “AI Assistant” upgrade. Eighty dollars a month per user, on top of what she was already paying. The pitch was that the AI would help her team draft tenant emails, summarize lease documents, and surface insights from her data.
She wanted to know if it was worth it. Twelve users at $80 a month is $11,520 a year. She wanted me to tell her yes or no.
I asked her what was actually killing her week. She said maintenance request triage. Tenants would call, text, email, and submit requests through the app. Three different inboxes. The same person would frequently report the same problem in two channels because she didn’t trust the first one was being read. Her team spent something like fifteen hours a week just figuring out which requests were duplicates and which were genuinely new.
The “AI Assistant” upgrade did not address that workflow at all. It would draft an email faster. It would summarize a PDF. Both of those were maybe five percent of what was eating her week. The actual chokepoint, the thing that was costing her real money, was untouched by the upgrade.
She paid $11,520 for the upgrade anyway, because the vendor made it sound like a strategic decision, and she felt behind. Three months later she still had the duplicate-request problem.
This is the bolt-on trap, and most small businesses are walking into it right now.
What’s happening in the market
Every SaaS vendor is bolting an AI layer onto their product right now. Salesforce has Einstein. HubSpot has Breeze. Microsoft has Copilot inside everything. Property management, dental practice management, legal case management, HVAC dispatch, and roughly two hundred vertical SaaS products have all rolled out an AI feature in the last eighteen months. The pricing is consistent. It’s an upcharge. Twenty to a hundred dollars per user per month.
The features are also consistent. They cluster around four things. Drafting emails. Summarizing documents. Generating reports from your existing data. A chatbot that answers questions about how to use the software you’re already paying for.
I am not saying these features are useless. Some of them are genuinely good. The Copilot summary of a long email thread is real value. A first draft of a tenant lease violation letter saves time. None of that is fake.
But here’s what those features are not doing. They are not fixing the workflow that’s actually slow inside your business. They are sitting on top of the existing software, doing thin work around the edges, and charging you a premium for the privilege.
Why vendors build it this way
The vendor is not stupid. They are doing the rational thing for their business model. They have a SKU. They have a per-seat license. They have a billing system designed for upsells. The cleanest way to capture AI revenue is to package it as another SKU and charge for it.
What they cannot easily do is rebuild the underlying workflow your business actually runs on. That would mean changing the data model, restructuring how their software handles your specific edge cases, and accepting that your operation does not work the same way as the average customer in their portfolio. None of that maps to a recurring revenue line they can sell to investors.
So you get the bolt-on. The bolt-on lives at the surface layer. It looks like AI. It costs like AI. But it does not change the way work moves through your business.
The honest test
Before you pay any vendor upcharge for an AI feature, ask three questions.
What workflow inside my business eats the most hours of paid labor right now? Write the answer down. Be specific. Not “operations.” Not “communications.” The actual thing. Maintenance request triage. Insurance verification. Quote preparation. Order reconciliation. Whatever it is.
Now look at the vendor’s AI feature list. Does the feature list include the workflow you wrote down? Not adjacent to it. Not “could help with it.” The actual thing.
If the answer is no, the upcharge is not solving your problem. It might still be worth $80 a month per user for the email drafting and the summary tool, depending on your team. That’s a different conversation. But you should have it as a different conversation, with clear eyes about what you are buying.
The trap is that vendors sell the upcharge as if it addresses the AI question for your whole business. It almost never does, because the workflow that’s costing you the most money is usually outside the boundaries of what the vendor’s product was originally built for.
What actually fixes those workflows
For the property manager, the answer was not inside her property-management software. It was a thin layer that sat between her three intake channels and routed the requests into a single deduplicated queue. The work involved reading from her email, her texting platform, and her tenant portal, running incoming messages through a classifier that grouped duplicates, and writing a single canonical record back into the system her team already used.
It was a $9,000 project. Not a recurring fee. One build. The vendor’s $11,520-a-year upcharge was a worse deal by a wide margin, and worse still, it did not even address the workflow it was implicitly being purchased to fix.
This pattern is structural. The cheapest, most useful AI work inside a small business almost always sits in the cracks between your software systems. It connects two things that don’t talk to each other. It cleans up data before it hits a tool you already use. It pre-processes a queue before a human touches it. None of that lives inside any single vendor’s product, because none of those vendors own the seam between your systems. You do.
The seam is where the value is. The vendor cannot sell you the seam, so they sell you the bolt-on instead.
Where this gets uncomfortable
I want to be honest about a piece of this that does not flatter my own business model.
There are vendors right now doing real work to embed AI more deeply into their core product. A handful of them are rebuilding their data layer, their workflow engine, and their automation tooling so that AI is not a bolt-on but a structural part of how the software runs. When that happens well, it is genuinely better than anything an outside firm can build for you, because the vendor has access to data and surfaces that an external integration does not.
The honest answer if you find one of those vendors covering your industry is to stick with them. Do not hire someone like me to build a parallel system. Use what you have. Save the money.
The catch is that those vendors are rare, and most of the AI upgrades you are being pitched right now are not them. They are the bolt-on. So the work, if you are serious about it, is to tell which is which before you pay.
A simple way to ask
If you are evaluating an AI upgrade from a vendor you already pay, ask the salesperson one question. Show them the workflow that’s costing you the most money inside your business and ask them to walk you through, click by click, how their AI feature changes that specific workflow.
If they can do it, the upgrade is probably real for you. Pay it.
If the answer is a generic explanation about how the AI “works across your data” or “augments your team,” the upgrade is a bolt-on. It might still be worth something. It is not worth what they are charging.
That conversation takes ten minutes and it usually saves a small business several thousand dollars a year. It is the single most useful thing I would tell a small business owner to do this quarter, and it does not involve hiring me at all.
Which, I think, is the point.