The promise
When AI coding tools took off, a lot of people and companies figured they could just build it themselves now. Describe what you want, and the AI writes it.
And for a while, it feels like that's true. You get something on the screen fast. It looks close.
Where it gets stuck
The first 80% of a project is the part AI makes easy. The last 20% is where it gets hard: security, handling real data, edge cases, deployment, performance, and keeping the thing running when something breaks at 2 a.m.
That's the part where you need to know what to ask for, and how to tell when the AI got it wrong. Without a software background, it's really hard to know what you don't know.
The tools amplify experience
A non-engineer with AI tools becomes something like a 1x engineer. An experienced engineer with the same tools becomes 10x, sometimes 100x, because they spend their time on architecture, reviewing, and judgment while the AI handles the volume.
It's not about whether you use AI. It's about what you bring to it.
What that means if you're deciding
If you have someone on your team with real software experience, give them AI tools and get out of their way. If you don't, the fastest and cheapest path is usually one experienced person who can build the whole thing, rather than a team or a stalled DIY project.
Either way, I'm happy to give you an honest take. Send me a message.