For the last decade, "build vs buy" was a question mostly reserved for funded startups and enterprise IT teams. Everyone else just bought SaaS — often paying for 80% of features they'd never use, locked into annual contracts, hoping the provider wouldn't pivot or raise prices. That's changing. Fast.

What is vibe coding, actually?

Vibe coding is a term for the practice of building software primarily through conversation with an AI — describing what you want in plain English, having the AI write the code, then iterating on it until it does what you need. No computer science degree required. No learning syntax. You're the product manager; the AI is the developer.

The tools that make this possible have matured dramatically. In 2026, the main players are:

• Claude Code / Cursor — AI-powered coding environments where you describe features and the AI writes, edits, and debugs the code. Best for building complete applications.
• v0 by Vercel — Generates frontend UI components from a text description. Good for quickly building the visual layer of a tool.
• Bolt.new — Full-stack apps from a prompt, deployed in minutes. Lowest barrier to entry for simple tools.
• Replit — Browser-based environment with AI assistance. Good if you want to iterate quickly without setting up a local development environment.

These aren't toys anymore. They're production-capable tools that can ship real software — databases, authentication, integrations, APIs — without the person running the conversation being a programmer.

Why it matters for small business owners

The economics are stark. A typical small business owner in NZ or Australia might pay:

Most of these tools were built for a median customer, not for your specific business. You pay for features you never use, and you're missing the ones you actually need. With vibe coding, the alternative is now plausible: describe your exact workflow, and build a tool that fits it precisely — for a fraction of the recurring cost.

What does a session actually look like?

Here's a realistic example. A small electrical contracting business in Auckland was using three separate tools — one for scheduling, one for invoicing, and one for tracking materials used on each job. None of them talked to each other. They were exporting CSVs and re-entering data manually.

Over a weekend, the business owner sat down with Claude Code and described the problem:

"I need a simple tool where I can log a job, add the materials we used, mark it as done, and auto-generate an invoice I can email to the customer. I also want to see a weekly summary of which jobs we completed and how much we invoiced."

Within a few hours of back-and-forth, they had a working prototype with a database, a form for job entry, a materials list, and PDF invoice generation. A few more sessions to refine it, and they had replaced two of their three tools. Total ongoing cost: the hosting, roughly $10–15/month.

That's not hypothetical. That's what's happening now, in real businesses, with AI tools that didn't exist two years ago.

What vibe coding is good for

• Internal tools — dashboards, job trackers, material cost calculators, timesheet systems. Anything used internally doesn't need to scale to millions of users.
• Replacing one specific feature from an expensive tool — if you're paying $200/month for a platform but only use two features, build just those two features.
• Automating a manual process — weekly reports, invoice reminders, importing data from one system to another.
• Prototyping a product idea — before spending on development, validate whether the concept works with a rough build.
• Industry-specific calculators or tools — charge-out rate calculators, compliance checklists, estimate generators tailored to your trade.

Where it breaks down

Vibe coding has real limits. It's important to understand what it can't easily do:

• Complex integrations at scale — connecting to Xero, Stripe, or other APIs isn't impossible, but it requires more back-and-forth and some understanding of what you're wiring together.
• Security-critical systems — anything handling payments, personal data, or medical records at scale should be reviewed by an actual developer before going live.
• Maintenance over time — code you don't understand is harder to maintain. If the person who built the tool leaves or the AI session context gets lost, the next update is harder.
• Complex multi-user systems — tools used by one or two people are much simpler to build than tools with complex permission systems, concurrent users, and audit trails.

The mental model to carry: vibe coding is best for your problem, not for a product you're selling to thousands of others. The more bespoke and internal, the better it works.

The bigger shift: from "I can't build this" to "let me try"

The psychological barrier matters as much as the technical one. For most of the history of software, if you weren't a developer, your options were: buy something that roughly fits, hire someone expensive to build something, or do it manually. The middle ground didn't exist.

That middle ground now exists. It's imperfect and it takes more patience than clicking around in a SaaS dashboard — but the ability to say "I have a specific problem and I'm going to build a specific tool for it" is now genuinely within reach for non-technical people. The skills that matter most aren't programming skills. They're problem clarity, iteration patience, and the willingness to try.

If you're a business owner spending $300+ a month on software that doesn't quite fit your workflow, it's worth asking whether 2026 is the year you try building something instead.