Table of Contents
The number that should worry your SaaS vendors
Start with the number everyone quoted. Retool's 2026 Build vs. Buy report found that 35% of teams have already replaced at least one SaaS tool with software they built themselves. And 78% expect to build more internal tools this year.
Here's the honest caveat, because it matters. Retool sells a tool for building software, and the report surveyed 817 of its own customers and builders. That sample leans toward people who already like to build. Discount the raw number if you want to.
The direction holds up anyway, because it lines up with two things that are harder to spin: what other analysts are seeing, and what public-market investors are pricing into software stocks right now. The same report found 60% of respondents had built software outside IT's oversight in the past year, and 25% do it often. That's not a tooling trend. That's people routing around procurement because the gap between what they were sold and what they need got too wide to live with.
The signal underneath the stat: when a quarter of your staff is quietly building software off the books, the question isn't whether to allow it. It's whether you'll govern it or pretend it isn't happening.
What they replaced first
The interesting part isn't that companies are building. It's what they chose to build first. Retool's respondents ranked the tools they replaced:
| What got replaced | Share who replaced it |
|---|---|
| Workflow automations | 35% |
| Internal admin tools | 33% |
| BI and dashboards | 29% |
| CRMs and form builders | 25% |
| Project management | 23% |
| Customer support tools | 21% |
Notice what's missing from that list. Nobody is rebuilding email. Nobody is writing their own payroll system. The tools getting replaced are the operational, internal, glue-layer systems where a generic product never quite fit the way the business actually runs.
That's the pattern worth holding onto. Companies replaced the software where the distance between "what the SaaS does" and "what we actually do" was widest. The closer a tool sits to how your company is genuinely different from your competitors, the worse a one-size-fits-all subscription serves it.
Why now: AI changed the build math
For fifteen years, build-vs-buy was settled by one fact. Building was slow and expensive. Buying was fast. So you bought, and you bent your process to fit the tool, and you paid the subscription forever. The math wasn't close.
AI moved the line. A workflow tool that used to take a small team a full quarter to build now takes a couple of weeks. When the cost of building drops by that much, every "we'll just buy it" decision you made under the old math is worth reopening. That's the whole story behind the Retool number. It isn't that companies suddenly love building. It's that building stopped being the expensive option.
One warning, because this is where projects go wrong. AI did not make building free. It made the first 70% fast and left the last 30% exactly as hard as it always was.
The last 30% is integration, edge cases, security, and the unglamorous work that decides whether a tool survives contact with production. AI accelerates the demo. It does not finish the system. A prototype that works on Tuesday and a tool a hundred people depend on for payroll are separated by exactly that 30%, and it still takes real engineers to cross it.
So the build math changed, but it didn't disappear. The companies that win here aren't the ones that vibe-coded a replacement over a weekend. They're the ones that used AI to build the right things fast and still treated the result like production software.
The SaaS economics quietly coming apart
There's a second force pushing companies to build. It has nothing to do with your team. It's happening inside your vendors' own business models, and it's getting worse.
Per-seat pricing was built for a world of human users who log in. Bessemer's 2026 AI Pricing Playbook found pure per-seat pricing fell from 21% to 15% of vendors in a single year, while hybrid models jumped from 27% to 41%. The reason is structural: AI agents don't log in and don't map to headcount, so the seat stops representing value. A vendor whose product replaces ten people can't keep charging for ten seats.
At the same time, AI features are expensive to run. ICONIQ's 2026 State of AI report put average gross margins for AI products near 52%, well below the 75% to 85% that classic SaaS enjoyed. Inference, model hosting, and orchestration all scale with usage instead of sitting at near-zero marginal cost.
Why this lands on your desk
Your vendors are caught between repricing, which alienates customers, and absorbing inference costs, which crushes margins. Neither is your problem to solve. But it's why the vendor you renew with this year may look very different next year: new pricing, new packaging, new owners. Early in 2026, both Fortune and TechCrunch ran "SaaSpocalypse" headlines as investors repriced the big SaaS names. The tremors are real, and they reach your renewal terms.
Where the build-vs-buy line actually sits now
None of this means rip out your SaaS stack. That's the overcorrection, and it's just as expensive as the one it replaces. The honest answer is that the line moved, not that it disappeared. Here's a cleaner way to draw it.
| Keep buying | Worth building |
|---|---|
| Email, calendar, payroll, accounting, identity | The workflow nobody else runs the way you run it |
| Commodity tools where a vendor will always do it cheaper | Systems built around data you own and want to keep owning |
| Anything that doesn't make you different from a competitor | The integration glue between systems that vendors won't connect |
| Tools you use straight out of the box, unmodified | Tools you've already bent, scripted around, and fought with |
The test is three questions. Does this tool make you measurably different from your competitors? Do you own the data inside it, and would you rather not hand it to a vendor? Is the gap between what the SaaS does and what you actually need wide enough that you're already working around it? Three yeses, and you're looking at something worth owning.
That last row is the tell most companies miss. If a team has already wrapped a SaaS tool in scripts, exports, and manual workarounds, they've effectively been maintaining custom software for years. They're just doing it on top of a subscription they also pay for.
What to do before your next renewal
You don't need a strategy offsite for this. You need an afternoon and your list of contracts. That's it.
Pull every SaaS renewal coming up in the next twelve months. Flag the ones your teams have customized heavily or bolted automation onto, because that's the build signal hiding in plain sight. Then ask one question per tool: if this got three times more expensive at your next renewal, would you pay it, or would you build? The tools where the answer is "build" are the ones to look at now, before the renewal forces a rushed decision under someone else's deadline.
And here's the part the subscription model never gives you. When you build it, you own it. No seat count. No annual increase you didn't negotiate. No vendor repricing your roadmap because their margins moved. The software does what your business does, and it keeps doing it whether the vendor pivots, gets acquired, or sunsets the product you depend on. That's not a cost decision. It's a control decision, and 35% of companies have already started making it.
Looking at a SaaS tool you'd rather own?
Codavyn builds custom software for enterprise and government teams, accelerated by AI and engineered for production. We help you decide what's worth building, then build it on a fixed scope and a clear timeline. You own the result outright.
Not sure where the line sits for your stack? Start with an honest read on what to build and what to keep buying. See our AI Readiness approach →
