Let me tell you about two feature requests.
The first one was filed on November 19, 2017. Seven years ago. The ask: let an admin change a customer’s email address in Jira Service Management. Not migrate accounts through a four-step workaround involving Atlassian ID. Just… change an email address. The kind of thing you’d expect a junior developer to ship on a Tuesday afternoon. As of today, it has 779 votes, 447 watchers, and a status of “Future Consideration.” Which, in SaaS-speak, translates roughly to: “we heard you, we filed it, please stop asking.”
The second one landed in January 2019. The request: make Confluence’s alphabetical page sorting persistent, so new pages don’t just pile up at the bottom like uninvited guests. Six years later: 600 votes, 261 watchers, status “Under Consideration.” Progress!
Now, to be fair to Atlassian, they’re not uniquely terrible. Every major SaaS vendor has a graveyard of feature requests exactly like these. Sensible, obvious, clearly wanted by thousands of paying customers. Just… never quite prioritized. Because they have a roadmap. And you’re not on it.
This is the part of the SaaS brochure they don’t show you.
The Pitch vs. The Reality
SaaS vendors are exceptionally good at one thing before you sign: making you feel like you’re about to get exactly what you need. The demos are polished. The slide decks are beautiful. The onboarding is smooth. Eighty percent of your requirements? Covered, on day one.
It’s that remaining twenty percent where things get interesting.
That twenty percent is where your actual workflows live. The edge cases. The things specific to how your organization actually operates. And when you file a support ticket asking about them, you enter a fascinating parallel universe where time moves differently. Features are “on the roadmap.” Updates will come “in a future release.” Your vote has been registered. Thank you for your feedback.
Meanwhile, you’re paying. Every month. For the product as it exists, not as it was promised.
The Numbers Are Telling
The scale of SaaS sprawl is difficult to overstate. According to BetterCloud’s annual State of SaaSOps report, the average number of SaaS applications per company peaked at 130 in 2022 and even after a wave of consolidation, still sits at over 100 today. That’s more than 100 subscriptions to manage, renew, secure, and integrate. For every single organization.
The waste embedded in that sprawl is just as striking. Gartner estimates that approximately 30% of purchased SaaS licenses go unused, what they bluntly call “toxic spend.” BetterCloud puts a price tag on it: companies report wasting an average of more than $135,000 per year on unused software licenses alone. And Gartner projects SaaS spending will continue growing at around 19% annually, reliably outpacing the budgets meant to fund it.
Do the arithmetic on a typical mid-market tool. Fifty users. €50 per user per month. That’s €30,000 per year. Every year. With annual price increases that arrive in your renewal email as a polite fait accompli. After five years, you’ve spent €150,000-plus on software you don’t own, can’t modify, and can’t easily leave and somewhere between a quarter and a third of those licenses have been sitting idle.
Something Changed
Here’s what’s different in 2026: building software got dramatically cheaper and faster. Not incrementally but by an order of magnitude.
Tools like Claude Code and Kiro represent a new category of agentic coding assistants. They don’t just suggest the next line. They can take a requirement, reason through a solution, write the code, test it, catch errors, and iterate, with minimal human supervision. What previously required a team of developers and months of work can now be done by one technically capable person in days.
This isn’t science fiction. It’s happening right now in engineering teams across the world.
And it fundamentally changes the math on one of the oldest questions in IT: should we build or buy?
What You Get When You Build Your Own
When you build a custom application -even a relatively simple internal tool- you build exactly what you need. No more, no less. You control the data model, the workflow, the integrations, and the roadmap. And critically: you decide what gets built next. Not a product manager in Sydney who’s never seen your workflows.
Running on cloud-native infrastructure like AWS means scalability, security, and availability are largely handled by the platform. The operational gap between “something you built” and “something a vendor hosts for you” has narrowed considerably. And the cost gap has flipped.
A custom-built equivalent to that €30,000/year SaaS tool, developed with AI-assisted tooling and running on serverless AWS infrastructure, might cost a fraction of that annually in operational expenses, after a one-time build investment that has also come down dramatically. More importantly: you own the asset. It does exactly what you need. And you’re not waiting seven years for someone to let you update an email address.
A Word of Honesty
Custom software isn’t free of responsibility. Someone has to maintain it, secure it, and evolve it. The SaaS argument, that someone else handles the infrastructure, the uptime and the patches, is not without merit. And for commodity functions like email, video conferencing, or payroll, that argument still wins. These are solved problems. Building your own would be an expensive act of reinvention.
But pairing custom-built applications with managed cloud services gives you the operational coverage of SaaS without the product dependency. You get uptime. You get security. You get to decide what comes next.
So What Should You Actually Do?
Not everything should be custom-built. That would be its own kind of madness.
The question worth asking is simpler: where are your core differentiating workflows? The processes that encode years of operational knowledge. The edge cases that keep showing up in feature request threads, yours and everyone else’s, year after year, status unchanged.
Those are exactly the features that will never quite make it onto a SaaS vendor’s roadmap.
The question is no longer “can we afford to build?” The question, in 2026, is whether you can afford to keep waiting.
JSDCLOUD-5746 would like a word.
