SaaS is dead. Not the delivery model — software will keep showing up over HTTPS. What’s dead is the deal: pay forever to mold your business around someone else’s product, then pay again to bolt on the parts they didn’t build.
For twenty years that trade made sense. Building custom was expensive, slow, and risky. Buying NetSuite or Salesforce was the safe call — even if “safe” meant your finance team learned to live with seven workarounds and your sales process bent itself around objects named for somebody else’s industry.
That math has flipped. And the companies still pretending it hasn’t are about to lose a decade of operating leverage to the ones who notice.
The compromise tax
Every off-the-shelf ERP or CRM ships with the same hidden line item: the compromise tax. It doesn’t appear on the invoice. It shows up as:
- A finance team running parallel spreadsheets because the GL structure doesn’t match how the business actually thinks about margin.
- A sales ops hire whose entire job is keeping Salesforce in sync with reality.
- An implementation partner billing six figures to make the platform behave like the brochure said it already did.
- Quarterly “upgrades” that break the integrations you paid the implementation partner to build.
- Workflows your team has memorized but cannot explain to a new hire, because they only exist as folklore around the platform’s limitations.
Add it up. For a mid-sized business, the compromise tax is routinely larger than the license fee itself — and you pay it every year forever, with no asset to show for it.
The pitch was always: “Don’t reinvent the wheel.” The reality was: “Buy our wheel, then pay us to make it round.”
What changed
Three things, all in the last 24 months:
1. AI collapsed the cost of custom software. A small team augmented with modern LLM-assisted engineering ships features in days that used to take quarters. The economics of bespoke have moved roughly an order of magnitude.
2. The integration layer got cheap. Stripe, Plaid, modern ERPs-as-APIs, embedded analytics — the boring plumbing that used to eat 60% of a custom build is now a weekend’s wiring.
3. The biggest players already admit it. Palantir’s entire pitch — Foundry, AIP, Apollo — is that generic SaaS doesn’t work for serious operations. They sell an enterprise operating system, not a packaged app, because they figured out that the world has fragmented into “a sky full of clouds” and rigid SaaS can’t span it. When the most valuable software company of the cycle is telling enterprises that off-the-shelf doesn’t fit, it’s worth listening.
The headline writes itself: the era of renting someone else’s idea of your business is over.
What “fits your business” actually looks like
A real custom ERP isn’t a Salesforce clone with your logo on it. It’s the system where:
- The objects in the database are the objects your team already talks about — not “Opportunities” and “Accounts” but production runs, batches, trial lots, whatever your industry actually uses.
- The workflows mirror how decisions are actually made, not how a generic platform’s product manager guessed they should be.
- The reports show the numbers your CFO cares about, not the dimensions some BI vendor preselected.
- Integrations exist where you need them and nowhere else. No SOC2-flavored sprawl of “tools we technically pay for.”
- When the business changes, the software changes — in days, by your team. Not in a $400k re-implementation cycle.
We just built one of these. A small-goods manufacturer was running their entire operation across QuickBooks, three spreadsheets, a shared inbox, and tribal knowledge. They’d been quoted six-figure implementations for NetSuite that would have replaced four of those tools with one tool that didn’t quite fit any of them. We built them a bespoke ERP — production planning, inventory, batch traceability, costing, the works — wired exactly to how they actually run the business. It cost a fraction of the NetSuite quote, ships changes the same week they ask for them, and the team uses it because it speaks their language.
That’s the move. That’s what’s available now.
”But what about [thing the platform does]?”
The standard objection: Salesforce/NetSuite/HubSpot does a thousand things. You’d have to rebuild all of that.
No. You’d have to build the dozen things you actually use. The other 988 features are why the platform is slow, expensive, and confusing. They’re the reason your team needs training. They’re the reason consultants exist. You’re not getting value from them; you’re getting taxed by them.
A custom system is small on purpose. That’s the feature.
”Isn’t custom software a liability?”
It was. When it was hand-rolled PHP in 2009, maintained by one person who eventually left — yes. That story scared a generation of executives into renting forever.
Modern custom isn’t that. It’s typed, tested, version-controlled, documented, infrastructure-as-code, and increasingly co-maintained with AI. The maintenance profile of a well-built modern custom system in 2026 is closer to “a well-maintained npm dependency” than “the legacy ERP that haunts the basement.”
Meanwhile, your SaaS vendor’s roadmap is a liability you don’t control, priced in perpetuity, with a contract clause that lets them raise it 7% a year.
Which is the real liability?
How to actually do this
If you’re sitting on a multi-year SaaS contract bleeding budget into workarounds, the playbook is:
- Map the compromise tax. Add up the workarounds, the spreadsheets, the integration costs, the implementation retainers, the headcount that exists because the platform doesn’t fit. Be honest. It’s bigger than you think.
- Identify the dozen workflows that actually matter. Not the 988 features the platform ships. The dozen things your team does every day where the platform makes them miserable.
- Prototype the replacement, fast. A focused team can stand up a working version of the core in weeks, not quarters. The point is to prove the economics, not to boil the ocean on day one.
- Migrate by surface, not by big bang. Replace one workflow. Then the next. The SaaS sits alongside until it’s been hollowed out, then it gets cancelled.
We do this work. We’ve shipped production AI systems, bespoke trading platforms, domain-specific SaaS, ML systems wired into real business decisions, and computer vision platforms for clients who’d been quoted “buy the platform” by everyone else. The bespoke ERP for the small-goods manufacturer is the same playbook applied to a problem the SaaS industry pretends is theirs alone.
It’s not. It’s yours. Build it.
Stop renting compromises. Let’s talk about what your business should actually run on.