Soft ComputersA new survey reported by ZDNet found that 80% of developers admit AI coding tools feel more addictive than helpful. They keep reaching for tools like GitHub Copilot, Cursor, and ChatGPT to generate code, not because the output is always good, but because it feels fast. That gap between feeling productive and actually being productive is something every Canadian SMB paying for custom software development should understand right now.
What the Survey Actually Found
According to ZDNet's August 2026 reporting, the survey results point to a pattern where developers lean on AI code generation heavily, sometimes without fully reviewing what the tool produces. The tools autocomplete functions, suggest entire blocks of logic, and can scaffold a working prototype in minutes. The problem is that "working" and "well-built" are not the same thing.
Code that looks fine on the surface can carry security gaps, skip proper error handling, or create technical debt that costs far more to fix later than it would have cost to write it properly the first time. When a developer is moving fast and accepting AI suggestions without careful review, that risk increases.
Why This Matters If You Are an SMB Hiring a Developer
If you have hired a freelancer, an agency, or a small dev shop to build something for your business, whether that is an e-commerce site, a client portal, an internal tool, or a mobile app, you are probably not watching what tools they use or how carefully they review AI-generated output. Most project contracts do not address this at all.
This creates a real risk. You could receive a delivered product that passes a basic demo but has problems underneath: unvalidated form inputs that expose you to injection attacks, hardcoded credentials left in by an AI suggestion nobody caught, or logic errors that only show up when a specific edge case hits your real data.
We are not saying AI coding tools are bad. We use them ourselves and they do speed up legitimate work when applied carefully. The issue is the word "carefully." Not every developer or shop using these tools is doing the review work that responsible use requires.
What to Ask Before You Sign a Development Contract
You do not need to become a technical expert to protect yourself. You need to ask the right questions before the project starts.
- Do you use AI code generation tools? There is no wrong answer here, but you want to know. A developer who says yes and then explains their review process is more trustworthy than one who says no and clearly is.
- What does your code review process look like? Any shop delivering production software should have a defined review step, whether that is a second developer reviewing pull requests or a structured testing phase before handoff.
- Will you provide a code audit or security scan before delivery? Tools like Snyk, SonarQube, or even GitHub's built-in security scanning can catch common vulnerabilities automatically. Ask if this is part of their process.
- Who owns the code at delivery, and can we get it reviewed by a third party? If they are resistant to an independent review, that is a signal.
- What is your warranty or support period after launch? Bugs that come from rushed AI-generated code often surface in the first 30 to 90 days of real use. Make sure the contract covers fixes in that window.
What We Do Differently
At Soft Computers, we do use AI coding tools on development projects. We are direct about that. What we also do is treat AI output as a first draft, not a finished product. Every function generated by an AI assistant goes through a human review before it makes it into a client's codebase. We run static analysis on every project. We test edge cases. We do not ship something because it worked in the demo.
The addictive quality the ZDNet survey is describing is real. It is easy to get pulled into moving fast because the tools reward speed with something that looks like progress. Slowing down to review, test, and validate is the discipline that separates software that holds up in production from software that creates a support nightmare six weeks after launch.
If You Already Have Software Built, Here Is What to Do
If you have had a website, app, or internal tool built in the last 12 to 18 months and you are not sure how much AI-generated code is in it or whether it was properly reviewed, a code audit is worth doing before something breaks in production.
We offer code reviews for existing projects. We look at security posture, code quality, and whether the architecture is going to scale with your business or become a problem as you grow. It is a fixed-scope engagement and you get a written report you can act on.
Reach out to us directly if you want to know what is actually inside the software your business runs on.