A client asked me something last month that I have been getting a lot lately. She had just talked to a developer about her project, and he mentioned, almost in passing, that he uses AI tools to help write code. She called me right after, worried. "Does that mean he's cutting corners? Am I paying full price for a project that's basically auto-generated?"
I get why she asked. "AI wrote the code" sounds like it could mean anything, from a helpful shortcut to a red flag. So let me answer it plainly, as a developer who uses these tools every day: no, it doesn't mean corners are being cut. But the question underneath — how is the AI actually being used — is exactly the right one to ask. Here's what I'd want you to know before you ask it.
What "AI-Assisted" Actually Means Day to Day
When people picture AI writing software, they often imagine typing a request into a chatbot and getting a finished app back, untouched by human hands. That is not how professional developers work, and it is not how I work.
In practice, tools like GitHub Copilot, Claude, and Cursor sit inside the same editor a developer already uses. As I type, the tool suggests the next few lines of code, the way your phone suggests the next word in a text message. I can accept the suggestion, edit it, or ignore it completely. Nothing ships until I have read it, tested it, and decided it belongs in the project.
Think of it less like a robot builder and more like a very fast, very well-read assistant. It has seen an enormous amount of code before, so it recognizes common patterns quickly. But it doesn't know your business, your customers, or what "done" means for your specific project. That part is still on me.
This distinction matters more than it sounds. A developer typing with AI assistance is still making every decision about what the software does and why. A tool that generates an entire application with no one checking the output is a completely different situation, and a much riskier one. Almost nobody serious builds real business software that way, but it is worth asking your developer which one describes their process.
Where AI Genuinely Saves Time
I won't pretend these tools don't help, because they genuinely do, and pretending otherwise wouldn't be honest. The place they help most is repetitive, well-understood work: setting up a standard form, writing the boilerplate that connects a webpage to a database, generating test cases for code that already exists, or converting a block of code from one format to another.
This is the coding equivalent of a contractor using a pre-cut set of studs instead of measuring and cutting each one by hand. The measuring skill still matters. But not measuring the same board fifty times a day frees up hours for the parts of the job that actually require a trained eye.
For my clients, that speed shows up as lower cost and faster delivery on the routine parts of a project — the login screen, the basic CRUD page, the confirmation email. It does not show up as a shortcut on the parts that make your software actually work for your business.
Where Human Judgment Still Does the Heavy Lifting
Here is the part I think gets lost in the "AI wrote it" headline: the hardest, most valuable decisions in a software project are not about writing lines of code at all. They are about architecture, business logic, and security — and none of those are things you hand to an AI tool and walk away from.
Architecture is the set of decisions about how all the pieces of your software fit together, and how well it will hold up as your business grows. That takes judgment built from having watched systems succeed and fail over years. Business logic is the specific set of rules that make your software yours: how a discount gets applied, what happens when a customer cancels mid-order, which staff member is allowed to approve a refund. An AI tool has no way to know these rules unless a person who understands your business writes them in, checks them, and tests every edge case.
Security is the sharpest example. AI-generated code can absolutely contain mistakes that expose customer data or leave a door open for attackers, the same way code written entirely by hand can. The tool doesn't know your compliance requirements or what happens if a mistake reaches production. A developer who understands the risks has to review that code specifically with those risks in mind, every time.
A Concrete Example
Say you're building an online booking system for a small clinic. AI assistance might speed up writing the calendar widget, the confirmation email template, and the basic form validation — routine pieces with well-known patterns.
But deciding what happens when two patients try to book the same slot at the same second, how far in advance a cancellation still gets refunded, and how patient records stay private and compliant — that requires someone who understands your clinic's actual rules and the regulations around health data. No tool suggests that logic on its own. A developer designs it, writes it deliberately, and tests it against the scenarios that matter to your business. I've written more about how AI is changing software development more broadly, if you want the bigger picture.
Questions Worth Asking Your Developer
You don't need to understand the technology to ask good questions about it. Here are four I'd encourage you to ask any developer or agency:
- "Do you review and test AI-suggested code before it ships, or does it go out untouched?" The answer should always be that a person reviews it.
- "Where do you use AI tools, and where do you avoid them?" A thoughtful answer will mention things like security-sensitive code or core business logic as areas that get extra human attention.
- "Who is accountable if something goes wrong?" The answer should be a person or a company, never "the AI."
- "Can you walk me through how a key feature actually works?" A developer who understands the code, AI-assisted or not, can explain it in plain terms. If they can't, that's the real warning sign — not whether they used a tool.
The presence of AI in a developer's toolkit isn't the risk. The absence of a person who reviews, tests, and takes responsibility for the result is. That's true whether the code was typed by hand in 2015 or suggested by an AI tool this morning.
If you're evaluating a developer or agency for your next project, I've also written a guide on how to choose the right software development partner that covers the bigger vetting questions beyond just AI use.
I use these tools myself, every day, because they make the routine parts of my job faster. What I don't do is let them make the decisions that matter for a client's business. That line is the one worth asking about — not whether AI was involved at all.