For fifteen years, "buy, don't build" was settled wisdom. Renting software from Salesforce was obviously cheaper, faster and safer than building your own — and for most companies, it was. That assumption is now worth re-examining. Not because SaaS got worse, but because the cost of building software collapsed. AI-assisted development has quietly changed the one variable the entire SaaS business model rests on: the price of writing and maintaining your own code.

This piece walks through the actual numbers for a German SME, comparing a Salesforce subscription against a bespoke CRM built and maintained by a small, AI-augmented team. The conclusion isn't "SaaS is dead." It's more useful than that: for a growing set of companies, owning your core software is now cheaper over five years than renting it — and the crossover point is arriving earlier every year.

The subscription you actually pay for

Salesforce's list prices (2026, per user per month, billed annually) run: Starter €23, Pro Suite €92, Enterprise €162, Unlimited €322. Enterprise is the realistic tier for any serious sales team, because forecasting, quoting, workflow automation and full API access are gated there.

But the sticker price is the smallest part of the story. Three things compound it:

Price increases are structural, not occasional. Salesforce raised list prices ~9% in 2023 (its first in seven years) and another ~6% in August 2025. Most contracts also embed an 8–10% annual uplift clause on your contracted rate — so real renewals routinely land at 9–18% year-over-year before you negotiate. Whatever you pay today, model it rising every single year.

Add-ons carry the margin. The base licence is the entry ticket. Real deployments accumulate AI credits (Agentforce), integration middleware (MuleSoft, often €90,000+ for enterprise integration), Premier Support (30% of net licence fees), extra sandboxes, storage and CPQ. Add-ons commonly represent 25–40% of contract value.

Someone has to run it. Beyond ~10 users, Salesforce effectively requires a dedicated or fractional administrator. In Germany, that's a €59,000–€69,000/year role. Even a half-time admin adds €25,000–€35,000 fully loaded — every year, forever.

None of this is hidden or scandalous. It's simply the true cost of the subscription model: you are renting, and the rent goes up.

What "build" costs now that AI does the heavy lifting

The old objection to building was effort. A custom CRM covering the essentials — contacts and accounts, pipeline and deal tracking, activity logging, email integration, reporting dashboards, roles and permissions, GDPR-compliant data handling — used to mean 8–14 weeks of a full agency team at Western rates, and a long tail of maintenance nobody wanted to own.

AI-driven development changes both halves of that equation.

The build. A lean, AI-augmented team — in practice, a project builder (architect/tech lead) plus a senior loop engineer running an agentic "harness and loop" workflow — can take a greenfield application from prototype to production dramatically faster than a traditional team. GitHub's controlled study found developers completed a task 55% faster with an AI pair programmer. Across the industry, well-scoped greenfield work is being compressed by an estimated 30–55%. A CRM that was a 600-hour build in 2022 becomes a 250–350-hour build today. "Vibe coding" gets you a working prototype in days; disciplined loop engineering hardens it into something you'd trust with customer data.

A realistic figure for a production-grade operational CRM built this way lands around €60,000–€100,000 as a one-time cost.

The maintenance. The industry rule of thumb is 15–25% of build cost per year — roughly €10,000–€18,000 annually for bug fixes, security patches, dependency updates and new features, handled by the same small team on a part-time retainer. Hosting is the pleasant surprise: an EU-based, GDPR-compliant production setup (for example on German or Finnish infrastructure) runs €50–€120/month — €600–€1,440/year — for 10 to 50 users. You own the code, the data and the roadmap.

The five-year comparison

Figures are planning ranges, rounded, with Salesforce at Enterprise pricing plus modest add-ons and an ~8%/year renewal uplift applied. Treat them as a model to run against your own quote, not gospel.

10-seat SME:

50-seat mid-market:

The shape of the result is the important part. At 10 seats, the two options run close over three years, with the build pulling ahead by year four or five. At 50 seats it isn't close: Salesforce becomes three to four times more expensive over five years. The reason is structural — your build cost is fixed, while subscription cost scales linearly with every new seat and compounds with every annual uplift. The more you grow, the worse renting looks.

Why this matters more in Germany

There's a second dividend that has nothing to do with cost. A self-hosted CRM on EU infrastructure sidesteps the entire transatlantic data-transfer question — the US CLOUD Act, the recurring "Schrems" legal challenges to the EU-US Data Privacy Framework, and the uncertainty a German company inherits when its customer data lives on US SaaS. Building on EU-hosted infrastructure means data residency and GDPR posture are design choices you control, not terms you accept. For a German SME with a compliance-conscious board or customer base, that's a strategic asset, not just a line item.

The caveats

This is not a blanket "everyone should build" argument, and the math only works under real conditions.

• Execution risk is the whole game. The savings are real if the build is tightly scoped and well managed. Underfund maintenance and you accumulate technical debt that erases the advantage.
• AI productivity claims are genuinely mixed. The headline speed-ups come largely from vendor-adjacent research; a rigorous 2025 study (METR) found experienced developers were actually slower with AI on mature, complex codebases. Greenfield CRM work is far more favourable — but don't bank on a specific multiplier. Pilot it.
• You inherit responsibility. Security, uptime, compliance engineering and continuity become yours. Salesforce's ecosystem, governance and AppExchange are real value for complex, multi-department operations.
• Small and simple still favours SaaS. If you're a 10-seat team with vanilla needs, a lighter SaaS tool (HubSpot, Pipedrive, Zoho) may beat both on speed and risk. The build case gets compelling at roughly 25+ seats, or three or more genuinely custom workflows, or a hard data-sovereignty requirement.

The takeaway isn't that SaaS is finished. It's that the default has flipped from "obviously buy" to "genuinely worth modelling" — and for scaling, compliance-sensitive companies, building and owning is increasingly the cheaper path.

And you can push the economics further

Everything above assumes a competent, competitively-priced senior engineering team. If you go one step further and combine AI-driven development with talent arbitrage — assembling that lean team through senior engineering hubs such as Vietnam, where a project-builder and senior loop engineer cost a fraction of German or Western rates — the build and maintenance figures compress toward the low end of every range above, and often below it. The AI does the productivity work; the arbitrage does the rate work. Stack the two and a 50-seat company's five-year saving over Salesforce can widen well past 75%. It's the difference between a good decision and an obvious one.

Before your next Salesforce renewal, it's worth running your own five-year numbers — with the annual uplifts, the admin cost and the add-ons all in — against what a lean AI-driven build would actually cost you. If you'd like a second pair of eyes on that model, or a scoped pilot to pressure-test the build option before committing, that's a conversation worth having early rather than at renewal time.