An Era Is Ending
The SaaS industry dominated software procurement from roughly 2010 to 2025 because it solved a real problem. Building custom software was expensive, slow, and required specialized talent most companies didn't have. Renting software from a vendor was the rational default.
That era is ending. Not slowly — quickly. And the companies that notice first will compound a durable advantage over the next decade.
I'm not saying SaaS is dying. It isn't. Payment processors, email deliverability, authentication providers — the true infrastructure layer of the web — will stay SaaS forever. But the layer above that, the productivity and workflow tools most knowledge-work companies spend 3–6% of revenue on, is in the middle of a structural shift.
We analyzed 340 B2B software procurement decisions in 2025 across our consulting book and industry data. The pattern was unmistakable: for the first time since 2010, custom software became the cheaper, faster option for a meaningful percentage of common business workflows.
This piece explains what changed, what it means, and what happens next.
The Three Collapses That Made Custom Viable
Three independent trends converged between 2023 and 2025. Any one would have been interesting. All three together changed the math.
Collapse #1: Build cost fell 80%
In 2019, building a custom CRM for a 50-person team cost roughly $180,000 and took 4–6 months. We have the invoices to prove it — we built exactly this for a client in 2019.
In 2025, we built the same functional CRM for a similar client in 14 days for $22,500.
What changed isn't just "developers use AI tools now." It's a three-layer productivity multiplication:
- AI coding assistants (Cursor, GitHub Copilot, Claude Code) generate 60–70% of the routine code
- Production-ready component libraries for common patterns (tables, forms, charts, auth) skip the boilerplate
- Full-stack frameworks (Next.js, tRPC, Prisma) eliminate the integration layer that used to consume weeks
Each layer compounds. What took 6 months now takes 2–4 weeks, and the quality is comparable or better because senior engineers spend their time on architecture and edge cases instead of typing.
Collapse #2: SaaS prices kept rising
While build costs fell 80%, SaaS prices did the opposite. The Vertice SaaS Pricing Index shows an average 11.4% annual price increase for B2B SaaS from 2020 through 2025. Some categories are far worse:
- Project management tools: +14.2% YoY average
- Support desk software: +18.1% YoY average
- Customer success platforms: +16.8% YoY average
- Sales tech stack: +13.6% YoY average
Compound that over 5 years and the average SaaS tool costs 1.7x what it did in 2020. Over 10 years, it's 2.9x. Every renewal season has become a budget negotiation.
Collapse #3: Lock-in got worse
As SaaS vendors faced pricing pressure from below (custom) and competition from alongside (other SaaS), they responded by tightening lock-in. More proprietary data formats. More AI features locked to the highest tier. Longer contract commitments. More integration entanglement.
We audited 15 popular SaaS tools and found that the 2025 version exports less usable data than the 2019 version did. "Data portability" marketing pages proliferated while actual portability regressed.
The combined effect: SaaS got more expensive AND harder to leave AT THE SAME TIME. That's when the math flipped.
The Data: What 340 Procurement Decisions Tell Us
Across our consulting and advisory work in 2025, we tracked 340 B2B software procurement decisions at companies ranging from 12 to 340 employees. For each, we recorded whether the team chose SaaS, custom, or hybrid.
The shift vs. 2022 baseline
| Decision Type | 2022 | 2025 | Change |
|---|---|---|---|
| Chose SaaS | 84% | 58% | -26 pts |
| Chose custom build | 5% | 29% | +24 pts |
| Chose hybrid (SaaS + custom modifications) | 11% | 13% | +2 pts |
The SaaS share dropped from 84% to 58% in three years. That's not a rounding error — that's a market shift.
Where the shift is concentrated
Not every category moved equally. The fastest replacement rates are in:
- Internal tools (dashboards, admin panels) — 78% now built custom
- CRM and sales operations — 41% now built custom
- Support desks — 38% now built custom
- Project management — 35% now built custom
- E-signature / document workflows — 33% now built custom
Where SaaS remains dominant:
- Payment processing — 98% still SaaS
- Email deliverability — 95% still SaaS
- Accounting / payroll — 87% still SaaS (regulatory moat)
- Authentication / SSO — 82% still SaaS
The pattern: commodity infrastructure stays SaaS. Workflow-specific tools go custom.
Why Vibe Coding Changes Who Can Build
"Vibe coding" — the practice of describing software in natural language and letting AI generate it — started as a developer-curiosity meme in 2024. By mid-2025 it had become a viable production methodology for well-scoped business applications.
I was skeptical. I changed my mind after watching three things:
1. Non-technical founders shipping production software. We onboarded a client last year — a 60-year-old insurance broker with zero coding background — who rebuilt his commissions tracking system using vibe coding tools in six weeks. The code was cleaner than what we would have produced ourselves, because the AI enforced consistent patterns he didn't know to violate.
2. Build speed for common patterns approaching zero. For a CRM, a booking system, a support inbox — anything with a well-established pattern — AI generation is now approaching the "type your request, get working code" threshold. Not for everything, not for novel algorithms, but for the 80% of business software that's workflow automation wrapped in CRUD.
3. The quality floor rising. Early AI-generated code in 2023 was buggy and needed heavy review. By late 2025, the baseline quality had risen to match what you'd get from a mid-level developer on their second project at a company. That's now the floor, and it keeps rising.
If your assumption about custom software is "that needs a $180K engineering project," your mental model is three years out of date. The real question isn't "can we afford a custom build?" It's "why haven't we already?"
What This Means for the Next Decade
A few predictions I feel confident making, ordered from near-certain to speculative:
Certain: SaaS pricing power will decline
Not uniformly — enterprise SaaS with regulatory moats stays priced. But mid-market productivity SaaS? The pricing power is gone. Expect 3–5 years of flat pricing or outright cuts as vendors try to retain mid-market accounts against custom alternatives.
Very likely: A new class of "build partner" emerges
Traditional dev shops charge $200K+ for custom builds with 3–6 month timelines. New AI-native build partners (AltStack is one example — there will be dozens more) compress that to $15–50K with 7–28 day timelines. This category is nascent in 2026. By 2028 it'll be a $20B+ industry.
Likely: Large SaaS vendors pivot to platforms
Retool, Airtable, and Notion are already pivoting toward "build your own tool" positioning rather than fighting custom head-on. Expect Salesforce, HubSpot, and Atlassian to launch versions of this strategy — essentially conceding the "pure SaaS" positioning and trying to capture the building-blocks layer.
Possible: Custom becomes the default for mid-market
Today it's 29% of procurement decisions. By 2028 it could be 50%+ for companies under 500 employees. Above that headcount, regulatory and scale concerns will keep SaaS dominant longer.
Speculative: A new generation of companies operates on mostly-owned software stacks
Some percentage of new companies founded in 2026+ will operate with custom software as the default and SaaS as the exception. Their cost structure will be structurally 40–60% lower than comparable companies built on SaaS stacks. This is a compounding advantage across a decade.
What You Should Do
If you're a founder, operator, or CTO reading this, the practical question is what to do right now.
One pilot this quarter. Pick your single highest-pain, highest-cost SaaS tool. Run a 2–4 week custom build pilot to replace it. Don't do all your tools at once — do one, learn the pattern, then decide if you want to expand.
Don't sign 3-year SaaS contracts. If a vendor pushes for a multi-year deal with a "discount," say no. Keep your annual renewal option open. The optionality is worth far more than the 5–10% discount.
Audit your stack annually. At renewal time for every SaaS tool, run the four-factor filter: Per-seat? Low utilization? Workflow-specific? Over $5K/year? If all four are yes, replace it.
Hire developers who can use AI tools effectively. The productivity gap between a developer who can effectively use AI coding tools and one who can't is now roughly 3–5x. This is the new baseline competency.
The Meta-Point
This isn't about "SaaS bad, custom good." It's about the fact that a foundational assumption — "custom software is too expensive to build" — that governed 15 years of purchasing decisions is no longer true.
When foundational assumptions change, entire industries reorganize. The companies that notice first, adjust, and act compound an advantage. The ones that don't notice keep writing larger and larger checks to vendors whose business model depends on them staying put.
The SaaS era happened because custom software was expensive. Custom software isn't expensive anymore. The era is ending.
The only real question is whether you're going to be early to notice, or late.
Ready to pilot a custom replacement? Book a free 30-minute call — we'll map your stack, identify the highest-ROI candidate, and give you a fixed-price quote to replace it in under 4 weeks.