LogoSmall
Login

An In-House App Built with AI: A Good Idea for a Rebar Company?

AI lets you build an app over a weekend. The real issue isn't AI: it's who secures it, who maintains it, and whether your company wants to become a software vendor.

MBBy Maxime Beaupré · Updated on · 6 min

With AI, anyone in a rebar company can now build an app over a weekend. That's not a problem in itself. The problem is what comes after: an app put online without real security, carrying all the flaws of any homegrown software, that someone in the company will have to maintain for years instead of doing their actual job. Before you dive in, the real question isn't "can AI do it?" but "do we want to become a software vendor?"

I've been programming for more than 30 years, and AI is a remarkable tool for people who know how to program. So I'm not going to tell you it's dangerous or that you should be wary of it. I'm going to tell you what it doesn't do for you. This article complements Excel, homegrown software or rebar software, which covers homegrown software in general.

What is a "vibe-coded" app?

The term vibe coding, which appeared in early 2025, describes a way of programming where you describe what you want to an AI, accept its code without really reading it, and keep asking until it works. Some platforms even put the app online, database included, in a few hours.

In a rebar company, a resourceful estimator or a computer-savvy detailer decides to replace the spreadsheet that's bursting at the seams. They ask the AI for an order form, a production board, a login for the shop foreman. On Monday it works, and everyone is impressed, rightly so.

Why is it so tempting?

Because it works, at least at first, and it's nearly free: no licence, no salesperson, no implementation project. The tool looks exactly like the company. AI really has lowered the barrier to entry, and for trying out an idea or automating a one-off calculation, it's an excellent tool.

But an app that works on Monday isn't software yet. Software still works three years from now, protects its data, and somebody knows how to fix it.

Is AI the problem?

No. Plenty of good programmers use AI every day. The difference lies in what surrounds the tool: someone who reviews the code, recognizes a security flaw, tests, backs up, and will still be around next year.

AI writes the code, but it doesn't take responsibility for the result. That falls on whoever puts the app online: for an in-house tool, your company.

What security risks can an app built in a weekend hide?

An app generated quickly does what it was asked to do, and people rarely ask "protect my data against someone trying to get in." The most common problems are well known:

  • No real authentication. A login page protects nothing if the data stays reachable without it.
  • No access control. Everyone sees everything, including your costs.
  • Exposed keys. Passwords and database access keys end up in the code or in the browser, within reach of anyone who knows where to look.
  • Components never updated. An app relies on dozens of components written by others, where flaws are discovered regularly.
  • No tested backup. Until someone has tried to restore the data, you don't know whether you can get it back.

These risks aren't theoretical. In a test published in 2025 by the security company Veracode, built around tasks designed to trigger four common types of security flaws, about 45% of the code produced by the AI models tested contained one of those flaws. The same year, a public vulnerability (CVE-2025-48757) was published for apps generated by an AI app-building platform: because access rules weren't properly configured, their databases could be read or modified without logging in. The platform's vendor disputes the vulnerability, on the grounds that each customer is responsible for protecting their app's data. That's exactly the point: whoever puts the app online is responsible for it.

In a rebar company, what's inside that app is your prices and margins, your customers, your project drawings, your employees' hours and rates.

The Canadian Centre for Cyber Security publishes baseline controls for small and medium organizations: patching, strong authentication, backups, access control, an incident plan. An in-house app on the internet is affected by every one of them. Who looks after that in your company?

Why do the old problems of homegrown software come back?

Because an app built with AI is still homegrown software: the tool has changed, not the situation. The problems described in our comparison of Excel, homegrown software and rebar software all apply, with a few twists.

Only one person understands it. Sometimes nobody does: when the code was accepted without being read, even the person who "built" it doesn't know how it works. The day they leave, the app becomes a black box.

The code can shift under your feet. A small change requested from the AI can rewrite what was working. Without tests, you find out when an invoice goes out wrong.

It doesn't know the trade. AI knows management apps in general, not the specifics of rebar: bars, bar sizes, weight per linear metre, bending shapes, bar lists, extras. It doesn't know that the shop fabricates from tags printed from the bar list, that those tags are scanned with a barcode reader in the shop and in the yard, or that heat numbers have to follow the steel. Nor that information has to flow from drawing to invoicing without being retyped. You'll have to explain all of it, one rule at a time.

It reproduces your habits, bad ones included, without bringing in what has proven itself in other fabrication shops.

Who is going to maintain it?

Nobody asks that question on the Monday the app works. Yet software that runs online needs maintenance that never stops:

  • applying security patches as soon as a flaw is published in a component;
  • checking that backups are happening, and that they can actually be restored;
  • keeping up with changes from the host and the services you use, which change their rules and their prices;
  • fixing what breaks, often in the middle of the busy season;
  • adding what's missing: a new type of contract, another format a customer requires, a new machine.

Someone in the company will have to do that work, and it's rarely a programmer hired for the job: it's an estimator, a detailer or the owner. Every hour spent fixing the app is an hour not spent estimating, detailing or serving customers. The cost doesn't show up on any invoice, but it gets paid all the same.

Software vendors exist for that reason: their teams do this work full time, for many companies that share the cost. You don't build your own benders or your own trucks. The question is the same for software: is that where you want to put your company's energy?

When does an in-house app built with AI make sense?

It has its place, and it would be a shame to pass it up, when:

  • it's used to try out an idea or show what you want before buying software;
  • it automates a one-off calculation, with no sensitive data;
  • it isn't open to the internet;
  • it could disappear tomorrow without the company grinding to a halt.

It becomes a risk when it holds your sensitive data, is open to the internet and the company depends on it.

What should you ask before putting an in-house app online?

  1. Who can log in, and who sees what once logged in?
  2. Where are the passwords and access keys stored?
  3. Who applies security patches, and how often?
  4. Where are the backups, and when did someone last try to restore them?
  5. If the person who built it leaves tomorrow, who takes over?
  6. How many hours a month does this app cost in maintenance, and what isn't that person doing in the meantime?

If several answers start with "I don't know," it isn't AI you need to question; it's the place the app has taken in the company. To compare with specialized software, the guide to choosing rebar software suggests the questions to ask a vendor, and the complete guide to rebar software gives an overview.

Metacis is rebar software built by people in the trade. If this question is coming up in your company, let's talk: request a demo.

Share

Frequently asked questions

Related articles

Metacis is rebar software designed by people in the trade.