Should You Build Your Own Quoting Software with AI? A CTO's View

More job shop owners are trying to build their own quoting software. With AI coding tools, it looks possible for anyone: describe what you want, watch the code appear, and in a weekend you have something that prices a part. We hear about it from customers and prospects every month.

Some of it is a great use of the tools. Some of it is a slow, expensive way to rebuild something you can rent for less than the cost of a few hours of your time each month.

We asked Bilal Tawfic, CTO of Tempus Tools, what he tells shops that are thinking about it. He uses AI coding tools every day, and he's careful about what he uses them for.

What AI coding tools are good at

AI coding tools work best when you already know what the result should look like. "When you know what you want the outcome to look like, you feed that into an AI agent, and then you can correct it and make sure it stays within the parameters where you want it to go," Bilal says.

They're also very good at prototypes: code you write to prove an idea works, knowing you'll throw it away. Software teams did this long before AI. The tools just make it much faster.

The problem is what happens when nobody steers. "When you let it loose, it doesn't really say no to anything. It just keeps saying yes," he says. "You can end up down rabbit holes."

It feels fast, but it costs time

Building with AI is fun, especially if you're not a programmer. You describe something you've wanted for years and watch it appear in front of you. It's easy to lose a weekend to it, then another.

Most people who've tried it, including people on our own team, recognise the pattern. You build a lot of small tools, enjoy every minute, and three months later you're not using most of them. They worked for a while, then got more complicated, and then got forgotten.

For a hobby, that's fine. For a small business, those hours are wages: usually the owner's, or a key estimator's. It feels quick because something appears on screen immediately, but getting from "it works for this part" to "the whole shop can rely on it" takes much longer.

Why quoting software is harder than it looks

Most shop owners know their own pricing inside out, so the requirements seem obvious. The difficulty is that quoting rules keep unfolding once you try to write them down.

Bilal uses a simple example. You want to quote a part on a sheet of material.

1. First, the part has to fit on the sheet. Don't quote it if it can't.
2. Then you remember you stock three sheet sizes. It fits on all three, so which one should the quote use?
3. The most efficient one. But what does efficient mean?
4. For one part, probably the smallest sheet. For 20 parts, two small sheets might be worse than one large sheet, so the answer changes with quantity.

"Natural language doesn't support it," Bilal says. "We use the word efficient, we throw around terminology that needs further definition and further definition."

That's one rule about sheet selection. A working quote also needs cut time, pierce counts, material prices, scrap, setup charges, processes like folding, welding and coating, minimum charges, customer pricing and more. Each one has its own exceptions.

Every new feature can break an old one

You can build this up a step at a time, and that's how real software gets built. But every addition risks breaking what's already there.

The AI won't stop you. Ask it to make sheet selection "more efficient" and it will change the code without necessarily caring what it did before. Professional software teams guard against this with automated tests and release processes that catch a change that quietly breaks last month's pricing. "These things were around before AI, but they're even more important now," Bilal says.

Without them, you find out a rule broke when a customer gets a strange price, or when a job loses money.

The AI makes decisions you don't see

Experienced engineers make dozens of trade-offs while building something, and they know why they made each one. AI tools make those trade-offs too, but they don't always explain them, and sometimes they don't ask. "Sometimes it will give you the option, but it won't be clear about what the trade-off is or what the decision will mean," Bilal says. "Sometimes it won't even ask you. It will just make a decision for you."

If you're not a software engineer, you won't know a decision was made until it causes a problem.

The maintenance bill nobody budgets for

Say you get it working. The ongoing cost is where most DIY software comes unstuck.

Dependencies need constant updates

Modern software is built on other people's code: frameworks, libraries and packages that solve common problems. Those packages are updated regularly, and they have their own dependencies, which are updated too. Updates bring performance improvements and security patches, and sometimes changes that break your code.

Security doesn't wait for you

Every so often, researchers find a serious vulnerability in a package that half the internet uses. Your software might rely on it seven layers down without you knowing. "You need to be aware of that and you need to make sure that you patch those things," Bilal says. If your quoting tool is online, holds customer details or accepts uploaded files, that matters a great deal.

The ground moves under you

Even if the code itself doesn't change, everything around it does. You update your browser, computer or server and the old version stops running. Cloud providers retire services and change how things work. Sometimes keeping software running means real rework, with no new features to show for it.

None of this goes away. A tool that runs fine for six weeks will eventually need someone to fix it.

Why quoting software is a special case

Some business software is fairly simple to build. Quoting for profile cutting and fabrication isn't, for three reasons.

CAD files

Customers send drawings in many formats. Some are open, like DXF and STEP. Many are proprietary and need paid licences to read properly, and none of them is a true standard. The CAD packages also release new versions constantly. For a few months only a handful of files arrive in the new version, then suddenly everyone is using it and your software needs to read it.

Bilal compares it to a gym membership. "The gym's bought all the expensive equipment. You pay your $20 a week and you get to use the expensive equipment whenever you want." A single shop building its own tool will usually be limited to the open formats, because licensing the rest doesn't make economic sense for one business.

Cut time

Working out how long a laser will take to cut a part depends on a lot of numbers and settings, and complex parts behave very differently from simple ones. Bilal's view is that a persistent shop owner could eventually get it working. But a quoting software company has processed millions of customer drawings, and all the edge cases that come with them. "Many years ago [we] ironed out many of the issues," he says. "Now we have a stable version that took us years to get to."

Nesting

How parts are arranged on a sheet drives material cost, and there are many nesting algorithms to choose from. Each one trades speed against accuracy. Choosing and tuning one is a project in its own right.

"But the big online cutting services built their own"

They did. The large online cutting services that let customers upload a file and get an instant price have built impressive software. They also have dedicated software teams, venture funding and years of work on geometry, pricing and nesting behind them.

A shop with one or two lasers and an owner coding at weekends is in a very different position. Having the same AI tools doesn't close the gap. The hard part was never typing the code.

The real cost comparison

Tempus Tools plans start from around US$75 a month. That includes reading and repairing CAD files, cut time, nesting, pricing rules for cutting, folding and other processes, quote documents, security patches, CAD format updates, hosting and a support team that will look at your drawings and answer your questions.

Compare that with doing it yourself:

The sums are simple. If your time is worth $40 an hour (it's more, trust us), two hours of fiddling, maintaining or poking around a month already costs more than a subscription. Most DIY tools need far more than that, and the hours tend to come out of the owner's evenings and weekends.

To put numbers on your current quoting process, try our quoting efficiency calculator.

What you should build with AI instead

None of this means job shops should stay away from AI coding tools. Bilal's advice is to keep experimenting: "It's good to be curious and see what's out there."

The best projects are the ones where you know exactly what the result should be and the rules don't change much:

• Integrations that move data between systems you already use
• Scripts that copy, rename or file documents you currently handle by hand
• Dashboards: quotes by status, orders this week, win rates
• Production screens: what's in the queue, what's due today

These are much easier to maintain than a quoting engine, because they don't involve pricing logic, geometry or nesting. They also work well alongside quoting software. Quote and order data from Tempus Tools can feed your own dashboards and production screens, and our integrations connect to accounting and other business systems.

If you're going to build anyway

If you decide to build something, Bilal has some practical advice:

• Know what you want before you start. The clearer the result in your head, the better the AI does.
• Watch for the decisions it makes. If it picks an approach, ask what the trade-offs are.
• Use version control from day one, so you can always go back. "If you feel like you're going down the wrong path, there's no harm in resetting back to where you were."
• Add tests, and question the AI agent if it changes them for any reason.
• Plan who will maintain it in a year, and what happens if they leave.
• Keep checking the cost-benefit. Count your hours honestly.

Use AI for what it's good at

AI coding tools are useful, and they're getting better fast. For integrations, scripts and dashboards they can save a small shop real time. Quoting is different. It sits on years of work on CAD files, cut times and nesting, plus a constant stream of updates and security patches. Rebuilding that yourself usually costs far more than it saves.

If your time is better spent winning work and running the floor, let someone else maintain the quoting software. See how Tempus Tools works, or compare plans.

Frequently asked questions

Can I build my own quoting software with AI?

You can build a basic prototype. Turning it into something the whole shop relies on is much harder: you need reliable cut-time and material calculations, support for customer CAD formats, testing so new features don't break old ones, and ongoing security and maintenance.

How much does it cost to maintain custom software?

The main cost is time. Dependencies need updating, security vulnerabilities need patching, and operating systems and cloud services change underneath you. For a small business, those hours usually come from the owner or a key staff member, and they continue for as long as the software is in use.

What should a job shop build with AI tools?

Projects with clear, stable requirements: integrations between existing systems, scripts that automate file handling, dashboards and production screens. Avoid rebuilding complex logic like pricing, nesting and CAD interpretation, where commercial software already carries years of work.

Why is quoting software for laser cutting hard to build?

It has to read many CAD formats, some of them proprietary, calculate cut times from part geometry and machine settings, nest parts efficiently on sheets, and apply pricing rules with many exceptions. Each of these takes significant time to get right and keep up to date.