Every enterprise software decision of the last twenty years was shaped by one piece of arithmetic: custom software cost ten times more than off-the-shelf and took ten times longer to arrive. Whatever the question, that maths gave the same answer — buy the product, bend the business.
So you bent. You adopted the “best practice” workflow that came in the box, even where your own process was the thing that made you better than your competitors. You paid per seat, per month, forever. You built the workaround spreadsheets. And you told yourself this was the sensible choice — because it was.
It isn’t any more.
What actually changed
AI-assisted delivery is not a tool that writes code while everyone crosses their fingers. It’s a change in how senior engineers work. The judgement — architecture, security, integration, what not to build — stays human. The keystrokes largely don’t.
The effect on the two numbers that always killed bespoke is blunt:
| Traditional bespoke | AI-assisted bespoke | |
|---|---|---|
| First working version | Months | Days |
| Production system | Quarters | Weeks |
| Cost | The reason you bought SaaS | 60–80% less |
When a working prototype arrives in days, the biggest risk in custom software — spending six figures to discover you specified the wrong thing — collapses. You see real software on your real problem before you commit to anything serious.
The SaaS bill nobody totals
Here’s an exercise that hurts. Take one significant SaaS product you run. Multiply the per-seat price by the seats, by twelve, by five years. Add the integration middleware, the premium tier you upgraded to for one feature, and the consultancy spend on making it almost fit.
For most mid-size and enterprise deployments that number comfortably exceeds the cost of building the thing properly, once — and at the end of the five years, you’d own the build. The SaaS spend, meanwhile, buys you year six at whatever the renewal letter says.
SaaS pricing was never the bargain; it was the financing plan for software you could’t afford to build. Now you can afford to build it.
What bespoke gets you that SaaS never will
Your process, kept. If your scheduling, pricing or logistics process is part of why you win, putting it through a generic product is unilateral disarmament.
Ownership. Your IP, your code, your data, your roadmap. No surprise repricing, no feature you rely on being deprecated, no vendor acquired by your competitor.
Integration as a feature, not a tax. Bespoke is built from your systems out, not bolted onto them.
A cost that ends. Builds conclude. Licences don’t.
Where SaaS still wins — and we’ll tell you
Commodity functions with no competitive edge in them — payroll, email, video calls — should stay bought, not built. If a product genuinely fits 95% of your need, buy it; we say so on discovery calls weekly. Bespoke wins where the fit is poor, the seat count is high, the process is the edge, or the integrations are the hard part.
That’s the honest boundary. What’s changed is where it sits: work that was “obviously SaaS” five years ago is now squarely in build territory, because building got an order of magnitude cheaper and faster.
The arithmetic that ran your last twenty years of software decisions has flipped. It’s worth thirty minutes to rerun it.