Every SaaS application I’ve ever built started the same way: someone was using a spreadsheet to do something a spreadsheet was never designed for.
A fitness instructor tracking 200 clients’ attendance and membership payments in a color-coded Excel file. A property manager juggling maintenance requests across 80 units in a shared Google Sheet that six people were editing simultaneously. A small manufacturer tracking job orders in a CSV they emailed back and forth.
These people weren’t behind the times. They were solving real problems with the tools they had. The spreadsheet worked — until it didn’t. Until it crashed. Until someone overwrote three weeks of data. Until it became too slow to open. Until they hired someone new who couldn’t figure out the system.
That’s the SaaS moment: when the spreadsheet breaks under its own weight and someone finally says “there has to be a better way.”
What AI Changed About Building SaaS Products
I’ve been building custom web applications for clients since around 2017. The process was slow. A simple but solid CRUD application — good data model, clean UI, proper authentication, basic reporting — took me six to eight weeks of solo work. That meant the economics only worked for clients with significant budgets or significant pain.
AI-assisted development changed the math. Not because AI writes the product for you — it doesn’t — but because it eliminates the parts of the project that aren’t actually hard.
Writing boilerplate user authentication from scratch: 3-4 days. Using Laravel Breeze and having AI help me customize it for a specific use case: one afternoon. Writing every CRUD controller and view by hand: a week. Having AI generate the patterns while I review and adjust: a day and a half. Scaffolding a basic API with proper validation and error handling: AI does a first draft in minutes that I’d spend hours on.
The time I saved isn’t time I stopped working — it’s time I redirected to the parts that actually matter: the data model, the user experience, the edge cases that will matter in six months.
The Parts of SaaS That AI Can’t Do
I want to be honest here because there’s a lot of hype about AI building your entire startup while you sleep. That’s not how it works.
Understanding the problem domain. Before I write a line of code for any SaaS project, I spend time understanding the business. What decisions are people making? Where does information get created? Where does it get consumed? What breaks down in the current system and why? AI can help you think through this, but it can’t do the discovery for you. You have to talk to humans.
Data modeling. The database schema you start with shapes everything downstream. Get it wrong and you’re refactoring under pressure six months later. I’ve seen AI suggest data models that look clean and break catastrophically when you try to add multi-tenancy or reporting. This is a place where experience matters more than speed.
Security. Authentication, authorization, input validation, SQL injection prevention, rate limiting — these are not places to let AI go unsupervised. AI-generated code in these areas is often directionally correct but missing the details that protect real user data. Review everything. Test everything.
What happens when it scales. A query that works fine with 100 records becomes a problem at 10,000 and a disaster at 100,000. AI doesn’t naturally think about this unless you ask — and even when you ask, you need to verify the answers.
A Real Example: ADHDCal
Earlier this year I built ADHDCal — a calendar and task management SaaS designed specifically for people with ADHD. I am one of those people. The existing tools weren’t built for how my brain works, so I built my own.
The full stack: Laravel back end, Vue.js front end, PostgreSQL, hosted on Cloudways. Custom calendar rendering, layered notification system, quick-capture task input, priority weighting system.
Before AI tooling, this project would have been two months of evenings and weekends minimum. I’m not a Vue.js expert — the calendar UI layer alone would have taken weeks of research and fighting documentation. With Cursor and Claude as my daily co-pilots, the whole thing was production-ready in about six weeks of part-time work.
The key: I used AI heavily for the parts I knew well enough to review (Laravel controllers, API design, testing) and even more heavily for the parts I didn’t (Vue component structure, front-end date handling). But I reviewed everything. I tested edge cases. I made the architectural decisions myself.
Is Your Business Ready for a SaaS?
Not every spreadsheet problem needs a custom application. Sometimes a better spreadsheet or an off-the-shelf tool is the right answer. But some problems genuinely need something custom — and those problems are more affordable to solve than they used to be.
Signs that a custom application might be right for you:
- You’re using a spreadsheet for something with more than 3 concurrent users
- You’ve outgrown multiple off-the-shelf tools and keep hitting their limitations
- Your “system” is actually 4 different tools that don’t talk to each other
- You’re spending significant employee time on manual data transfer between systems
- You have a workflow that’s genuinely unique to your business and no SaaS tool covers it
If one or more of those resonates, it might be time to talk. The economics are better than they’ve ever been. A solo developer with good AI tooling can deliver what used to require a small team.
In 2026, the question isn’t whether you can afford to build the thing. It’s whether you can afford to keep living in the spreadsheet.
If you’re at that inflection point — where your current tools are breaking under the weight of your business — let’s talk. I’ve helped businesses make this transition for 28 years and I’m not going to oversell you something you don’t need.