Every few years, the tech industry rolls out a new wave of tools that promise to kill custom software development. First it was “off-the-shelf software.” Then SaaS. Then no-code and low-code platforms. Now AI code generation.
Each wave was real. Each genuinely expanded what you could do without hiring developers. And yet the market for custom software keeps growing.
There’s a reason for that.
What AI Code Generation Actually Changes
Let’s be specific about what AI has changed.
AI coding tools have made developers significantly faster at implementation. Boilerplate that used to take a day takes an hour. Standard patterns get implemented consistently in minutes. Documentation generates alongside code instead of three months later when nobody remembers why we made that decision.
This is real and it matters.
What AI has not done:
- Reduced the need to understand your business deeply before building software for it
- Made architectural decisions easier or more reliable
- Eliminated the need for testing, security review, or production operations expertise
- Made one-size-fits-all software suddenly fit complex business needs
The building is faster. Figuring out what to build — and why — requires the same expertise it always did.
Why Complex Businesses Still Need Custom Software
The businesses that come to me for custom software share a common trait: their processes are specific enough, or their competitive edge is embedded deeply enough in how they operate, that general-purpose software doesn’t fit.
The manufacturing example: A manufacturer I work with has a production scheduling system that accounts for machine maintenance windows, operator certifications, material lead times, customer priority tiers, and a dozen other constraints specific to their operation. No off-the-shelf scheduling software models this correctly. The workarounds — spreadsheets, manual overrides, tribal knowledge — were costing them capacity.
Custom software that actually modeled their operation replaced all of that. ROI was clear within the first year.
The service business example: A professional services firm had billing rules that were straightforward to describe but too complex for any standard billing platform. Multiple rate structures, project-phase-based billing, client-specific discount structures, approval workflows — each one would’ve been an “advanced configuration” in a standard tool. The combination simply wasn’t supported.
Custom billing software that matched their actual process replaced a Frankenstein’s monster of partially-correct software and manual reconciliation.
The “Just Use AI Tools” Counter-Argument
I’ve heard the argument: AI is making custom software development so fast that it’s more competitive with SaaS than ever.
That’s actually an argument for custom software, not against it.
If AI tools mean I can build a system tailored to your exact needs in half the time it would’ve taken previously, the cost-benefit of custom software improves. The investment is lower; the fit is still perfect.
The comparison isn’t “custom software vs. AI-generated software.” Custom software can be AI-assisted. The real comparison is “software that fits your business vs. software your business has to adapt to.”
The No-Code / Low-Code Reality
I have nothing against no-code tools. For specific use cases — internal forms, simple automation, basic databases — they’re excellent. I use them myself.
The limit is complexity. No-code tools trade flexibility for ease. That’s the right trade for simple use cases. When your requirements exceed what the tool can model, you either simplify your requirements (which means compromising on what your business needs) or you start fighting the platform.
Fighting the platform is expensive and demoralizing. I’ve rescued projects that started in no-code tools and grew past the tool’s capability. The extraction is painful.
The Right Question
Instead of “can I solve this with off-the-shelf software?” the better question is: “What does it cost to fit my business to a general-purpose tool versus building a tool that fits my business?”
Costs of using a general-purpose tool:
- License fees (ongoing, often increasing)
- Workarounds for features it doesn’t support
- Training users on a tool that doesn’t match their mental model
- Integration complexity with your other systems
- The ceiling on optimization — you can’t make a general tool better for your specific case
Costs of custom software:
- Higher upfront investment
- Ongoing maintenance requirements
- Dependency on developer expertise
For many businesses, the math clearly favors custom. The ones where it doesn’t tend to have simple, standard processes where a general-purpose tool is a genuinely good fit.
If you’re trying to figure out whether custom software makes sense for your situation, let’s work through the question together. I’ll give you an honest answer even if that answer is “use the off-the-shelf tool.”