Anybody can build software now. Here is how to tell if it is real.

Two years ago, building a construction app took a team, a year, and money. Today one person with an AI coding tool can ship something that looks finished in a weekend.

Some of that is good. There is more software aimed at contractors right now than at any point in history, and some of it is genuinely built by people who have been on a roof.

But here's what changed: the demo no longer tells you anything.

A slick interface used to be evidence. It meant somebody spent real money and real time, which meant there was probably a real company behind it. That signal is dead. The interface is now the cheapest part. What costs money is everything you can't see on the sales call. Backups. Security. Support at 6am. A company that's still around in three years.

So contractors are being asked to bet payroll, job photos, customer lists, and pricing on companies they have no way to evaluate.

This is how you evaluate them.

Start here: the risk isn't that it's bad

You can survive bad software. It's annoying. You switch.

What you can't survive:

• Three years of job photos gone in the middle of a hail claim dispute
• Payroll doesn't run on Friday because the app is down and nobody answers the phone
• Your pricing and margins sitting in a system that got breached
• The company folds and your customer list goes with it
• You want to leave and find out you can't get your own data out

Every question below exists to protect against one of those five things. Not against a clunky interface.

Part 1: Is this a real company?

Ask these first. If a company can't clear this section, nothing else matters.

"How long have you been a company, and how many people work here?"

Good answer: A number. "We started in 2019, there are 34 of us, 6 in support."

Bad answer: Anything that avoids the number. "We're a lean team." "We're growing fast." "We keep it efficient."

Small is not the problem. Small and hiding it is.

"Send me your certificate of insurance. General liability and cyber."

A real company sends this within a few hours because their office manager pulls it up in ten seconds. You do this on every sub you hire. Hold your software company to the same bar.

Cyber liability is the one that matters here. If they hold your customer data and they get breached, that policy is what stands between you and the fallout.

Bad answer: "What do you need that for?"

"If you shut down tomorrow, what happens to my data?"

Almost nobody asks this, which is why the answer is so revealing.

Good answer: They've thought about it. Export tooling, a notice period, sometimes a written wind-down commitment.

Bad answer: Laughing it off. "We're not going anywhere."

Two-thirds of the software you're being pitched this year will not exist in 2029. That's not pessimism, that's the base rate. Ask anyway.

"Who founded this, and what did they do before?"

Good answer: Named people you can find on LinkedIn who worked in the trades, or who partnered with someone who did.

Bad answer: No names anywhere on the site. No company page. No address.

Then follow up with the real test:

Ask them a trade question

Ask what O&P is. Ask what a supplement is. Ask what happens on a job when the adjuster's scope comes in short.

If the person selling you construction software can't answer a basic question about construction, they didn't build it for you. They built it for investors, and you're the thing they're going to learn on.

Part 2: Whose data is it?

This is the section most contractors skip and later wish they hadn't.

"Do I own my photos and job records? Show me where it says that."

Don't accept a yes. Make them point at the clause in the terms of service.

Good answer: They send you the section. It's short and it says you own your content.

Bad answer: "Of course you do" with no document. Or a clause that grants them a broad, perpetual, sublicensable license to your content. That language shows up more than you'd think.

"Do you train AI on my data? Can I opt out?"

This one is new and it's the one to watch in 2026.

Your job photos, your customer names, your estimates, and your margins are all sitting in that system. If they're training on it, your pricing structure becomes part of a model that your competitor's tool might be running on next year.

Good answer: A clear yes or a clear no, in writing, plus an opt-out if it's a yes.

Bad answer: "We use AI to improve the product experience." That's a sentence that means nothing and it's designed that way.

Saying yes is not automatically a dealbreaker. Not saying is.

"Where is my data stored, and who else touches it?"

Every software company uses other companies underneath: hosting, storage, email, analytics. Those are subprocessors, and a serious company publishes the list.

Bad answer: They don't know what you're asking.

"Have you ever had a breach? What happened?"

Good answer: Either a straight no, or a straight yes with what they did about it. A company that handled a breach well and will talk about it is often safer than one that has never been tested.

Bad answer: A pause.

Part 3: Security, in plain language

You don't need to become an IT person. You need three answers.

"Do you have a SOC 2?"

SOC 2 is an outside firm auditing how a company handles data. It costs real money and takes months, which is exactly why it's a useful signal.

Here's the honest part: most small construction tech companies don't have one. That alone is not a dealbreaker, especially for a company under 20 people.

What matters is how they answer:

• Fine: "No. We're too small to justify it yet. Here's what we do instead, and here's when we plan to get one."
• Not fine: "What's SOC 2?"
• Not fine: Dodging, or implying they have one when they don't. Ask to see the report cover page and the audit date. If it's real, they'll show you.

"Does it have two-factor, and can I control who sees what?"

Two-factor login should exist. Not optional in 2026.

Permissions matter more than most contractors realize. In a lot of shops, everybody logs in under the owner's account, which means your new hire can see your margins, your customer list, and your pricing. Ask whether you can set roles. Ask whether you can see a log of who did what.

"Show me your status page."

A status page is a public page showing whether the system is up and what broke recently.

Good answer: A link, with twelve months of history on it, including outages. Outages on the page are a good sign. It means they report them.

Bad answer: No status page. That means either they don't track downtime, or they don't want you seeing it.

Part 4: Does it work where the work happens?

This is where most construction software actually fails, and it's the easiest part to test yourself.

Never evaluate a product on the salesperson's laptop. That laptop is on fast wifi, in an office, running the newest hardware, driven by somebody who uses the product every day. That environment does not exist on your jobs.

Do this instead:

The three-year-old phone test

Install it on your foreman's actual phone. Not your new iPhone. The scratched-up Android that's been in a truck for three years. If it's slow or it crashes there, it's going to be slow or crash on every job.

The airplane mode test

Turn off wifi and cell service. Do a normal task: photos, a form, a time entry. Turn service back on.

Did it save? Did it sync? Did anything get lost?

Half of the tools sold to contractors quietly fail this, and you find out on a job with no signal, holding a phone with 40 photos that never uploaded.

The worst-with-tech test

Hand it to the crew member who hates technology. Not your sharpest guy. Say nothing and watch.

If they can't finish the task without help, the rollout is going to fail no matter how good the product is. Adoption is the whole game. A perfect tool nobody uses is a line item on your P&L and nothing else.

Part 5: What happens when you want out?

Ask these before you sign, not after.

"Export my data. Right now, on this call."

Not a promise. Not a support ticket. Make them do it live while you watch.

What you're checking: Can you do it yourself without asking permission? Is there a fee? What comes out? A CSV of records is fine. A PDF of screenshots is not. Photos should come out at full resolution, not compressed thumbnails.

If you can't get your data out on your own, you don't own it. You're renting access to your own business.

"What are the cancellation terms?"

Get specific:

• Does it auto-renew, and how many days notice do you need to give?
• How much notice before a price increase?
• After you cancel, how long do you have to pull your data before it's deleted?

That last one catches people. Some tools cut access the day you cancel, and your export window closes with it.

Watch out for the annual contract on a new product

An established company asking for an annual commitment is normal. A twelve-month-old company asking you to lock in for a year, before they've proven they'll be around for a year, is asking you to take their risk for them.

Ask for month-to-month for the first 90 days. How they respond tells you a lot.

Red flags in the sales process itself

The way a company sells is a preview of how it supports.

• The discount that expires today. Real companies want you to succeed on the platform. Pressure means they need the close more than they need you to be a fit.
• They can't name one contractor you can call. Not a logo on a slide. A phone number of an actual owner who runs it.
• All the reviews were posted the same month. Check the dates. Check the app store. A wall of five stars from a two-week window is a campaign, not a track record.
• Demo environment only. If they won't let you touch a real account with real data, ask why.
• Every answer is "that's on the roadmap." Buy what exists today. Roadmaps are not contracts.
• The marketing says AI forty times and never says what it does. Ask what specifically it does, what it does when it's wrong, and how you'd know.

Test support before you buy

Send a support email at 7pm on a Tuesday. Not to your salesperson. To the actual support address.

Time the response. Support is never better after they have your money than it was while they were trying to get it.

The ten-minute version

If you do nothing else, do these five:

1. Ask for the certificate of insurance. Real company or not. 2. Make them export your data live on the call. Own it or rent it. 3. Ask if they train AI on your data, and get it in writing. 4. Airplane mode test on an old phone. Works in the field or doesn't. 5. Email support at 7pm before you sign. Time it.

Five questions. Ten minutes. They'll catch most of what's going to hurt you.

Why we're publishing this

Nobody has been checking. Contractors have been handed a market full of new software with no way to tell the serious companies from the ones that will be gone by spring.

That's what the DIALED Standard is: fifteen items a construction tech company should be able to clear before you bet your business on it. Published in the open, the same fifteen for everybody, so you can ask a vendor about any line on it and see how they answer.

The questions in this article are the same ones. You shouldn't need us to ask them for you.

Running something that passed this test? Or something that didn't?

Add your stack. Tell us what you run, what you dropped, and why. You'll see what contractors in your trade at your revenue are running.

DIALED Contractor. The record of what contractors actually run.