SaaS or custom-built: renting or owning your tools in Switzerland

For fifteen years, renting software by subscription was the default choice — and it was the right one. In 2026, three forces are shifting that trade-off: SaaS vendors' price inflation, Switzerland's tightening stance on data sovereignty, and the falling cost of code driven by AI development assistants. Owning your tools is becoming a serious option again, even for a ten-person SME. The question deserves to be asked afresh at every subscription renewal.

Why your software bill keeps rising every year

The software subscription won for good reasons. Immediate deployment, continuous updates, low entry cost: against in-house development that ran into months of delay and tens of thousands of francs, the match was settled. Every support function found its vendor — customer relations, project management, documents, signature, accounting — and every vendor found its per-user, per-month model.

The downside crept in more quietly. Among established vendors, raising prices on the installed base has become a growth lever in its own right: price-increase announcements keep coming, and the bill for a company of a few dozen employees climbs year after year while actual usage plateaus. Most teams use only a fraction of the functions they pay for.

This rent-seeking rests on a well-documented mechanism: vendor lock-in. The deeper a tool embeds itself in the organisation — proprietary formats, cross-integrations, multi-year contracts — the more expensive it becomes to leave, and the more the renewal negotiation tilts in the vendor's favour. This spending builds no asset. It buys, month after month, the right to keep accessing your own data.

I did the exercise for my own tax back-office before building FiscalDoc: market solutions ranged between CHF 25 and 60 per month per entity, i.e. CHF 600 to 1,400 per year for my two scopes — with, in most cases, storage at a US-based host.

What Swiss law changes in the equation

Second, more recent force: the regulatory framework. In November 2025, Privatim — the conference of Swiss data protection commissioners — published a resolution restricting public authorities' use of international cloud services for sensitive data, except where end-to-end encryption is used and the provider doesn't hold the keys^[1]. The position targets the public sector first. But regulated sectors — banking, insurance, healthcare — are progressively aligning with these standards, often pre-emptively.

The core of the problem lies in the extraterritorial reach of US law: data hosted by a company under US jurisdiction remains exposed to that law, even if physically stored in Switzerland. The revised FADP, in force since 1 September 2023, requires adequate technical and organisational measures^[2] — and a subscription with a vendor under foreign law doesn't mechanically guarantee them.

A custom-built tool hosted in Switzerland or Europe, whose code, encryption keys and operation the company controls, greatly reduces this category of risk. The scope becomes documentable. That's a property that a subscription, by design, rarely offers.

Does custom-built still cost what it used to cost?

Third force, the most structural in my view. AI development assistants have collapsed the time between a software idea and its functional implementation. A project the market billed five years ago at tens of thousands of francs and months of delay now ships on a timeline and budget that change the original arithmetic.

I verified this at my own scale. FiscalDoc, the local application that classifies my tax documents with an open-source model running on my Mac, took me three evenings. Zero subscription, zero data leaving my machine, and the CHF 600 to 1,400 in annual subscriptions avoided. The full account is in FiscalDoc, or replacing a SaaS with local AI. A tiny case, I admit: one user, a few hundred documents a year. But the mechanics it demonstrates hold at SME scale — describe your flows, build close to the need, own the result. That's precisely what I do in my custom-built solutions, with the same logic and additional production requirements.

Lowering the barrier doesn't mean everything should be built. It means the zone where custom-built becomes competitive has widened, and it will keep widening as these tools mature.

And who maintains the tool once the provider is gone?

That's the classic objection, and it's a legitimate one. A custom-built tool whose provider disappears, changes its rates, or loses its key developer recreates the very dependency one was trying to escape. This objection has long been enough to justify the subscription, despite its cost.

A third path has opened up over the past eighteen months: delivering the software along with its own assisted-maintenance tooling. The client receives the source code, in full ownership, accompanied by agents calibrated on that code's architecture and the organisation's business vocabulary — an agent that generates new features, an agent that documents changes, an agent that monitors the coherence of integrations. The client stops being a tenant, without becoming a prisoner of frozen code that would need to be handed to the next provider that comes along. They are autonomous, with reduced in-house skill or the partner of their choice.

This path has its conditions. The need must be sufficiently well-defined for a tool to be specified, and the organisation must have at least one person — internal or contracted — capable of engaging with this tooling. When these conditions are met, the previous decade's economic equation no longer holds.

Five questions before signing — or building

The choice isn't ideological. Some situations call for the subscription, others for development. Here's the framework I apply, in this order:

1. Does the tool touch your competitive advantage? What sets you apart deserves to be owned; a commodity can stay rented.
2. How many users will you be paying for in three years? Per-seat pricing becomes heavy with growth; custom-built has a fixed cost, independent of user count.
3. What data flows through the tool, and under what jurisdiction? Sensitive data or data covered by trade secrecy: sovereignty weighs heavily in the balance.
4. What's your time horizon? Over three years and beyond, the total cost of ownership of custom-built becomes comparable to, sometimes lower than, the cumulative cost of subscriptions. For a temporary need, the subscription wins.
5. Who adapted to whom? If your teams work around the tool with exports and spreadsheets, you're paying a hidden cost the bill never shows.

A single answer rarely settles it. All five together sketch out a well-founded decision — it's the framework I run through when scoping every custom-built project, before a single line of code is written.

Renting often remains the right decision

I'll say it plainly: standard accounting, basic payroll, an experimental six-month need call for no development at all. The subscription remains unbeatable there, and will stay that way. The shift described here doesn't lead back to the heavy IT of the 2000s; it widens the zone where the trade-off deserves analysis, instead of a habitual renewal.

The fundamental difference comes down to one sentence. Software built for the company and owned by it is an asset that sits on its balance sheet; a subscription is a recurring expense that never enriches the balance sheet. Over the long horizon, the gap is measurable — and in 2026 it's measured across a much wider zone than five years ago. To place this decision among an SME's other AI projects, I laid out a complete roadmap in What should a Swiss SME do about AI in 2026?.

Key takeaways

— In 2026, SaaS is no longer the default choice: it's a choice that must be justified, on the same footing as custom-built. — Three forces shift the trade-off: vendor price hikes, data sovereignty (FADP, Privatim's November 2025 resolution), falling cost of code. — Five questions settle it: competitive advantage, number of users, data sensitivity, time horizon, how much the tool bends the organisation out of shape.

FAQ

Isn't custom-built reserved for large companies? Not anymore. AI development assistants have dramatically lowered the entry cost: I built FiscalDoc alone, in three evenings, to replace CHF 600 to 1,400 in annual subscriptions. At SME scale, a targeted business tool is now scoped in weeks, not quarters.

What happens to my tool if the provider disappears? If you own the source code, the documentation and the maintenance tooling delivered with it, you can switch providers without losing the asset. That's the opposite of SaaS: when a vendor shuts down or removes a feature, all you're left with, at best, is a data export.

Does the FADP forbid me from using a US SaaS? No. It requires adequate protective measures, which remains possible with certain services and certain types of data. But Privatim's November 2025 resolution hardens the line for sensitive data, and regulated sectors are moving closer to it. Good practice is to qualify tool by tool, based on the data that flows through it.

Where do I start if I suspect I have one subscription too many? With an inventory. List your subscriptions, their real annual cost, and the share of functions actually used. Then apply this note's five questions to those combining high cost, narrow usage and sensitive data: that's usually where custom-built wins first.

Is your next subscription renewal coming up? The AI Usage Diagnostic: sixty minutes to map your real flows, identify what deserves a custom-built solution, what stays in SaaS, and what needs no AI at all. Book a diagnostic

Sources

[1] Privatim, Resolution on international cloud solutions, November 2025. www.privatim.ch/de/privatim-verabschiedet-resolution-zu-internationalen-cloud-losungen/ [↩]

[2] Federal Act on Data Protection (FADP), revision of 25 September 2020, in force since 1 September 2023. www.fedlex.admin.ch/eli/cc/2022/491/fr [↩]

Jérôme Deshaie is CEO and founder of MCVA Consulting SA, an augmented agency based in Valais. Fifteen years serving major international brands, now working directly with Swiss SMEs. Background.