AI Beyond Chatting: Build Real Apps by Conversation

Most business owners think of AI as something you chat with: ask a question, get an answer, done. That undersells what's happening right now in software development.

A newer generation of AI coding tools lets someone describe what they want in plain English, and the AI writes working code to build it. Not a mockup. Not a description of how you'd build it. Actual software you can click through and use.

We cover three of these tools in depth elsewhere: Cursor, Claude Code, and Codex. This post is about the bigger shift they represent, and where the line still is between "AI helped build this" and "AI needs a developer watching over its shoulder."

What Changed

A few years ago, building even a small internal tool meant a developer writing code line by line, informed by requirements you gave them. That's still how larger, serious projects work, and for good reason.

What's new is the middle ground. A developer can now sit down with an AI tool, describe a piece of software in conversation, watch the AI generate real working code, test it, ask for changes out loud ("make the search box filter as you type," or "add a column for last contacted date"), and have a usable tool within hours instead of weeks.

The AI is doing the typing. The developer is doing the thinking: deciding what the tool actually needs to do, checking the AI's work, catching mistakes, and making sure the result is something you can trust.

That combination is what makes this useful. AI alone, with nobody who understands software reviewing the output, produces things that look done but have gaps you won't notice until something breaks or leaks data. A developer alone, without these tools, is slower and more expensive for small, low-stakes projects than the same developer using them well.

Three Kinds of Small Business Apps This Makes Feasible

An internal inventory tracker. Say you manage equipment, parts, or rental gear and you're currently doing it in a spreadsheet that's gotten unwieldy. A simple internal app that lets staff check items in and out, search by name, and see what's running low is a realistic small project. Expect a working first version in a matter of days, used only by your team, with no public-facing risk.

A simple booking or scheduling tool. If your booking needs are straightforward, one calendar, one type of appointment, a form for customers to pick a time, this is very buildable, quickly. The moment you add things like payment processing, multiple staff calendars syncing with each other, text reminders, or stored customer contact details, the project grows. It's still doable, but it moves from a quick internal build to a real project that needs real testing.

A data cleanup utility. Many small businesses have a CRM, spreadsheet, or database full of duplicate entries, inconsistent formatting, or old records mixed in with current ones. A one-off tool that reads your data, flags duplicates or errors, and lets someone review and merge them is a great fit for this approach. It touches your real data, so a developer should still be the one running it and checking results before anything is deleted or merged permanently.

What's Realistic on a Small Budget, and What Isn't

Internal tools that only your team uses, don't touch sensitive customer data, and don't need to stay running under heavy load are the sweet spot. These can genuinely be built faster and cheaper than they could have been a couple of years ago. That's a real, practical benefit for a small business, not a sales pitch.

Here's where this doesn't shortcut the work: anything customers will use directly, anything that stores or processes personal or payment information, and anything with compliance requirements, such as health data, financial records, or industry-specific rules. Those projects still need proper planning, security review, testing, and ongoing maintenance, the same as before. AI tools speed up the writing of code. They don't replace the judgment needed to decide what the code should do, whether it's secure, or whether it will hold up over time.

The honest way to think about it: AI coding tools lower the cost of the first draft. They don't lower the cost of getting the important parts right. A developer who knows what they're doing, using these tools well, can offer you more for a smaller budget than was possible before. A tool with no developer behind it is a different, riskier thing entirely.

Where This Fits in Your AI Business Maturity

This shift touches two dimensions we look at in our AI Business Maturity Assessment: Technology and Talent.

On the Technology side, it means the cost of building small, custom internal software has genuinely dropped, which changes the math on projects you may have dismissed as not worth building custom. On the Talent side, it raises the value of having a developer who knows how to steer these tools responsibly, rather than someone experimenting with them unsupervised on anything that matters.

If you've got a spreadsheet held together with tape, a manual process eating up staff time, or an idea for a small internal tool you assumed was out of reach, it's worth a conversation. Contact us and we'll give you a straight answer on what's realistic, what it would take, and where the real risks are.