AI has collapsed the cost and timeline of custom software. What used to require six months and $350,000 can now be built in four weeks for a fraction of that price — and that changes the calculus for every mid-market operator sitting on a workflow held together with Excel and institutional memory.

There's a spreadsheet worth telling you about. It tracked every shipment of product moving from Asia into a distribution network — factories, freight forwarders, ports, customs, carriers. Dozens of data points per order, updated constantly, pulled from systems that didn't talk to each other. Someone had to go get that information, manually, and put it somewhere everyone could see it. That somewhere was Excel. The file was enormous. It had errors. It was always a few days behind.

What custom software used to cost

Several custom software development firms were asked to quote a transportation visibility tool: multi-system integration, a clean interface the team would actually use, automated exception flagging. Nothing exotic. The median estimate came back at $350,000 to build, over six months, with ongoing maintenance costs of approximately $100,000 per year. For a tool that replaced a spreadsheet. The business case was hard to make at that price. The spreadsheet stayed.

That calculus has changed completely.

The economics are not incrementally better. They are categorically different. What changed is that the bottleneck shifted from building to understanding. — the economics argument

What AI changed

That same transportation visibility tool — multi-system integration, live data, exception alerting, finance activities layered in — was built in four weeks. Not a prototype. Not a pilot. A working tool, fitted to the actual workflow, connected to the actual systems, used by the actual team. Ongoing cost: $7,500 per month. No upfront build fee. No six-month implementation. No change management project to get people to use a system that wasn't built for them.

AI collapsed the time and cost of writing software. Code that used to take senior engineers weeks to produce can now be generated, tested, and refined in days. The bottleneck shifted from building to understanding — and understanding, when you start from the workflow rather than a feature spec, is fast.

§ Key takeaways
  • The economics of custom software have inverted: what once required large engineering teams and 18-month timelines can now be built in weeks for a fraction of a SaaS platform's annual cost.
  • The real cost of generic SaaS is hidden: you pay licensing fees, then spend internal time bending your workflow to fit the tool, then accept results optimized for someone else's business, not yours.
  • The transportation tool case: a $350K software estimate became a $7.5K/month custom build in 4 weeks. That's not an outlier — it's what happens when you build from the workflow.
  • The window is narrow: organizations that move to custom AI now are building tools their competitors can't replicate from a SaaS catalog.
A team working around a laptop — the build in progress.
Four weeks. A working tool. The math has changed.

The real cost of the status quo

When the price of custom software was $350,000, keeping the spreadsheet was a reasonable decision. But the spreadsheet was never free. The team maintaining it spent meaningful hours every week populating, checking, and distributing that file — time that had a cost, not just in salary, but in what those people weren't doing instead. The decisions made on stale data. The workarounds that accumulated, became invisible, became the way things work. The adoption failure baked into software built for everyone but optimized for no one.

The question we ask at the start of every engagement: what would you build if you knew it would take four weeks and cost a fraction of what you've been quoted before? — the opening question

The quiet thesis

The old build-vs-buy framework — buying is cheaper and faster, building is better but expensive — is obsolete. Custom tools built with AI are now faster to deploy than most SaaS implementations. They cost less than the annual subscription fees of the platforms they replace. They fit your workflow because they were built from it. The organizations still defaulting to SaaS for every problem are not making a neutral choice. They are accepting a fit gap, a recurring cost, and a ceiling on what the tool can ever do for them.