A 55-person real estate investment firm in Atlanta used to pay for Salesforce, plus separate contracts with two real-estate software vendors, to run its operations. Then it built its own replacement using AI coding tools, at a cost of roughly $300 a month, and dropped the other contracts along with it. The firm’s partner told The Information the move saves the company about $100,000 a year. A 70-person professional rugby club in Seattle did something similar, replacing both its CRM and its ticketing system in four months. So did a 45-person startup in Utah, which swapped a $40,000-a-year Salesforce contract for a homemade tool expected to cost about $1,200 a year.

None of these are tech companies. They’re a real estate manager, a sports club, and a small software startup, ordinary businesses that decided the software they were renting no longer justified its price tag. Their decision is becoming a lot more common, and it’s worth understanding why, whether you’re running a 10-person shop or a 500-person operation.

Why the math changed

For two decades, the “buy” side of build-vs-buy usually won for anything beyond the largest enterprises. Building software in-house meant hiring developers, managing a multi-month project, and maintaining the result forever. Buying a subscription meant a predictable monthly fee and someone else’s engineering team doing the maintenance. That trade-off made sense when building was slow and expensive.

AI coding tools have narrowed that gap. A prototype that used to take a development team a quarter can now take days. Retool’s 2026 survey of over 800 business builders found that 35% of teams have already replaced at least one SaaS tool with something they built themselves, and 78% expect to build more custom tools this year. That’s not a fringe statistic anymore, it’s roughly a third of the businesses surveyed making a decision that would have looked reckless five years ago.

The pressure isn’t limited to small businesses, either. Sanofi, the French pharmaceutical company with roughly 75,000 employees, is routing 80% of its ServiceNow workload through AI agents built in-house, aiming to save at least $10 million a year, according to reporting from The Information. Research firm Gartner estimates that $234 billion of enterprise software spending, about one-fifth of the total, is now exposed to this kind of replacement by 2030.

Why “just build it yourself” isn’t the whole story

It would be a mistake to read this as “SaaS is dead, build everything in-house.” The same Retool survey found that security and compliance concerns are the second-biggest technical obstacle keeping AI-built tools out of production, cited by 41% of respondents. Only 8% of builders deploy AI-generated code without changes; most test it carefully first. And 60% of the custom software being built is happening outside formal IT oversight, which creates its own risk: tools that work great for one team but nobody can audit, secure, or maintain when the person who built them leaves.

Salesforce’s own president, Srini Tallapragada, put the vendor’s counterargument bluntly to investors, warning that companies “can’t vibe code your way to enterprise reliability and security.” He has a point worth taking seriously: workflow tools and admin dashboards are the easiest things to replace, but core systems that hold customer data, run compliance-sensitive processes, or connect to dozens of other systems are a different kind of project. Big software vendors aren’t standing still either. ServiceNow, SAP, and Workday have all introduced metering controls on how outside AI agents can access customer data, and several vendors are shifting toward usage-based pricing rather than flat per-seat fees.

Meanwhile, the cost of staying on the “buy” side keeps climbing. Microsoft raised Microsoft 365 and Office 365 prices as much as 14% for some enterprise plans effective July 1, 2026, on top of new usage-based charges for AI features once free capacity runs out. Rising subscription costs are exactly the kind of thing that makes a business owner start asking whether a custom alternative might pencil out.

What this means if you’re evaluating your own software stack

1. Ask which tools you’re paying for but barely using. The businesses in the examples above didn’t rebuild everything, they replaced the specific tool or workflow where a generic product was charging them for features they never touched. Before commissioning a custom build or renewing a subscription, get an honest read on what percentage of the tool’s functionality your team actually uses.

2. Ask what happens to the software after it ships. A partner who can build you something fast should also be able to tell you who maintains it, how security patches happen, and what your options are if your business needs change. A tool nobody owns after launch becomes a liability, not an asset.

3. Ask where the line is between “build” and “buy” for your specific business. Systems of record, the software holding your customer data, your financials, your compliance history, usually benefit from the maturity and support structure of an established platform. The workflows layered on top of those systems, the reporting, the internal tools, the process automations, are where custom building tends to pay off fastest.

The takeaway

The build-vs-buy calculus has genuinely shifted, and businesses that assume the old rules still apply are leaving savings on the table. But the businesses succeeding at this aren’t randomly replacing every subscription they can. They’re identifying specific, well-scoped tools where a custom build offers a clear advantage, and they’re building with the same discipline, security review, and maintenance planning that any serious software project requires.

If you’re weighing whether a piece of your software stack should be built, bought, or left alone, we’d be glad to talk it through. Email us at info@kodevox.com or reach out through our contact page.

Sources and further reading:

• PYMNTS — SMBs Swap Pricey SaaS Contracts for AI-Built Apps (8 July 2026)
• Retool — The Build vs. Buy Shift: AI, Shadow IT, and the SaaS Replacement Era (2026 Build vs. Buy Report)
• GTM360 Blog — SaaSpocalypse Now – Part 1 (22 July 2026)
• Gartner — Gartner Says $234 Billion in Enterprise Application Software Spend Is at Risk From Agentic AI (1 July 2026)
• US Cloud — Microsoft 365 Price Increase July 1, 2026: What Enterprise Buyers Need to Know

— The Kode Vox Team