Own or rent software? A decision guide for Danish businesses
SaaS subscriptions are easy to start and hard to get out of. Here is an honest review of owning vs. renting: the cost picture over time, data lock-in, the GDPR and exit perspective, and why AI agents have moved the line for what is realistic to own yourself. With a concrete decision framework.
Table of contents
Own or rent software? The answer depends on what the system means to you: rent standard systems far from the core business, but own (or self-host) the systems where your data and processes are the actual value.
The short answer: rent the systems that are commodities and far from your core business, and consider owning (or self-hosting) the systems where your data, your processes and your customisations are the actual value. For many years that answer was mostly theoretical, because "owning" required a development department or an expensive consultancy. AI-assisted development has changed that markedly: the cost of building and maintaining your own software has fallen so far that the question of owning is once again a real strategic choice, even for mid-sized Danish companies without their own IT team.
Owning software means running the system on an instance you control yourselves, with your data in your own database, without a per-user licence and with the right to change the system without asking a supplier for permission. Renting software is the opposite: a subscription (typically SaaS) in which the supplier runs the system, owns the infrastructure and sets the price, the features and the terms for getting out again.
This guide goes through both routes without a holy war. SaaS isn't evil, and owning isn't automatically wise. But the decision is often made by chance (someone on the team set up a trial account three years ago) and not as what it actually is: a long-term choice about costs, data and control.
Owning vs. renting: what does it actually mean?
The words are used loosely in the industry, so let's get the definitions in place before we discuss pros and cons.
Renting: the SaaS subscription
Software as a Service means that the supplier hosts and runs the software, and that you pay on an ongoing basis for access, typically per user per month. You own nothing beyond your data (and even that can be more complicated in practice, as we'll come back to). The advantages are well known: no servers to look after, updates happen automatically, and you can get started the same day. That's why SaaS won the past decade, and for many types of system it is still the right choice.
The rental model today covers everything from email and accounting to CRM, project management and HR. If you want a concrete picture of what a rented system can typically do in the sales domain, we've covered it thoroughly in our guide to what CRM automation actually is.
Owning: your own instance and your own code
Owning can mean several things, and it's worth distinguishing between them:
- Classic on-premise: software installed on your own servers with a one-off licence. That model is genuinely dying for new systems, but still exists in older ERP and specialist systems.
- Self-host: you run the software (often open source or fair source) on your own infrastructure or in a cloud you control yourselves. The code isn't necessarily yours, but the operation and the data are.
- In-house developed: software built for you: either by an agency, by your own developers or, increasingly, with AI agents as the build capacity. Here you own the code, the data and the direction.
What the ownership variants have in common: there's no meter ticking per user, no supplier who can change the terms unilaterally, and your data sits somewhere you decide yourselves.
Why the divide has become more important
Five years ago this was mostly a discussion for large companies. Two things have changed that. First, the SaaS portfolio in a typical company has grown year after year, so subscription costs taken together have become an item management actually looks at. Second, the alternative (building and running it yourself) has become radically cheaper with AI-assisted development. When both sides of the equation move, the sum has to be redone.
The real cost picture of SaaS subscriptions over time
The starting price of a SaaS product is rarely the problem. The problem is the curve the price follows as your usage grows and your dependence increases. Four mechanisms recur across virtually all subscription products.
The per-user price grows with the organisation
Most SaaS products are priced per user per month. It feels fair at the start: five salespeople, five licences. But the model means your software cost is coupled directly to your growth, every new employee makes the software more expensive, regardless of whether the system actually costs the supplier more to run. And because many systems increasingly have to be used by "everyone" (customer service needs to see the CRM, finance wants access to pipeline figures), companies often end up paying full licences for employees who only look in the system a few times a month. It's a pattern we've also described in our reviews of Salesforce alternatives and HubSpot alternatives: the licence model punishes exactly what the system ought to reward, the whole organisation using it.
Feature gates: features move up the tiers
The second mechanism is the tier logic. Features you regard as basic (API access, advanced reports, single sign-on, workflows) are typically placed in the more expensive tiers. And the tier boundaries move over time: a feature that sat in the middle tier when you signed may sit in the enterprise layer in the next version. The result is that the real price of what you actually need is often markedly higher than the tier you started on. The decision "let's just upgrade to get that one feature" is one of the most common ways the software budget grows without anyone actively having decided it.
Price increases you don't control
As a tenant you accept that the supplier can raise the price, and it happens regularly in the industry, typically justified by new features (in these years often AI features, which are added on top of the price whether or not you use them). The more deeply the system is integrated into your day-to-day, the weaker your position in that negotiation. The supplier knows it: the cost of moving is yours, not theirs. A price increase on a system the whole sales department lives in is almost always accepted, not because it is reasonable, but because the alternative is a migration project.
Data lock-in: the most expensive cost isn't on the invoice
The fourth mechanism is the most important, because it reinforces the other three: your data lives with the supplier, in the supplier's data model. The more time passes, the more history, the more integrations and the more habits are tied to precisely that system. Export is rarely impossible, but what you can export is typically raw data without the structure, automation and history that made the system valuable. Lock-in isn't a flaw in the SaaS model; it's part of the business model. The EU has also found it necessary to legislate in the area: the Data Act contains rules intended to make it easier for customers to switch between providers of data processing services, precisely because switching costs have in practice locked customers in.
None of the four mechanisms means SaaS is a bad choice. They mean that the price on the sticker isn't the price. The real cost picture is the sticker price plus growth coupling plus tier creep plus price increases plus the day you want out.
What you get by owning
The ownership side of the sum isn't just about avoiding the above. There are four positive gains that are hard to get in the rental model.
Your own instance, and your data with you
When the system runs on an instance you control, with your own database, the question "where is our data, and who has access?" is suddenly easy to answer. It has practical value every single day: you can run direct database queries, build your own reports, connect other systems without waiting for an API quota, and tell auditors and customers precisely where data is stored. And it has strategic value on the day something changes, a supplier closes, changes terms or is acquired. Your data doesn't move with the supplier's decisions, because it never lived with the supplier.
No per-user licence
If you own the system, employee number 50 costs the same as employee number 5: nothing extra. It sounds trivial, but it changes behaviour. In the rental model access is rationed ("does she really need a licence?"), and the system's value is limited by who can afford to be part of it. In the ownership model it's the other way round: the more people use the system, the more valuable it becomes, without the invoice growing. For companies that work seriously on bringing sales, operations and data together in one place (what we have called revenue operations in another context) this is exactly what is decisive: coherence requires everyone to be on board, and per-user licences are a tax on coherence.
Customisation without waiting for the supplier's roadmap
In a rented system your wishes are a "feature request" in a queue you can't see. In an owned system they are a task. If your business has a process that doesn't look like the standard template (and most companies more than a few years old do) you can build it into the system instead of bending the process to fit the software. This is where the difference between "we use a system" and "we have a system that fits us" arises.
Predictable finances
The ownership model has costs too (operation, maintenance, further development) but they are yours to manage and prioritise. There's no email saying the price is going up, or that the feature you depend on now sits in a more expensive tier. For the finance function, the difference between a variable, supplier-controlled cost and a fixed, self-controlled cost is often more important than the amount itself.
What used to make "build it yourself" unrealistic, and what AI agents have changed
If owning is so attractive, why does everyone rent? Because the ownership route historically had three hard prerequisites that few could meet.
Before: three barriers
The cost of building. Bespoke software required either your own developers or a consultancy, and both were expensive and slow. A system that in SaaS form cost a subscription cost, as in-house development, a project with a requirements specification, rounds of tenders and months of development, with a real risk that the result still wouldn't hit the need.
The maintenance burden. Software isn't finished when it's built. Security updates, bug fixes, new needs, integrations that change: everything the SaaS supplier handles invisibly had to be staffed yourself. Many in-house systems ended up as "the system only Torben understands", and when Torben left, the system became a problem instead of an asset.
The skills gap. Even with money for both, it took technical leadership to commission, assess and run software development. Without a person who could translate between business and code, in-house development became a gamble.
Together, the three barriers made the rental model the only sensible choice for most, not because renting was best, but because owning was out of reach.
Now: AI agents as build capacity
That is the equation AI-assisted development has changed. Modern AI agents can today write, change and test code at a pace and a price that were unthinkable a few years ago. That doesn't mean software builds itself, or that you can do without judgement, but it moves the line for what a company without a development department can realistically own:
- The cost of building falls, because a large part of the actual coding work can be done by AI agents under human direction. What used to require a team can often be run by a few people who describe the need precisely and quality-assure the result.
- Maintenance becomes a conversation instead of a project. "Add a field", "make a report on X", "integrate with Y" are tasks an AI agent can solve in the system you own, without a ticket queue at a supplier.
- The skills gap narrows, because the requirements shift from "can write code" to "can describe the business precisely and judge whether the result is right". That is a skill many business people already have.
It is precisely this development that we are building Agent360 OS on: a Danish platform where every company is meant to be able to get its own instance; it has not been released yet. The point here isn't the product, but the principle: when build capacity becomes available, the question of owning becomes relevant for companies that have never had it as a real option.
An honest caveat is in order: AI agents don't remove the need for architecture, security thinking and operations. They remove the bottleneck on the building work itself. Companies that think owning is now free are simply making the classic in-house development mistake with new tools. More on the pitfalls at the end.
The middle way: fair source and self-host
Owning or renting is fortunately not a binary choice. Between the two extremes there are models that combine the convenience of the rental model with the control of the ownership model.
Self-host: rented code, owned operation
Many mature software products (especially open source) can be self-hosted: you run the software itself on infrastructure you control. You don't write the code yourselves, but your data sits with you, and nobody can shut off access or raise the price of something running on your own machines. The price is operational responsibility: updates, backup and security are your table. For companies with even a little technical capacity (or a partner who supplies it), self-host is often the best compromise for systems with sensitive data.
Fair source: open code with a time delay
A newer middle form is fair source licences: the supplier runs and develops the product commercially, but the code is available, and after a defined period it becomes genuine open source. That changes the balance of power fundamentally compared with classic SaaS: you can read the code, you can self-host, and even in the worst conceivable scenario (the supplier closes or changes course) the code becomes free, and you (or anyone else) can carry it on. For a customer, it means that "what happens if the supplier disappears?" has a concrete answer in the licence, not just a promise in a sales presentation.
The hybrid in practice
Most companies sensibly end up with a portfolio: rented standard systems for the generic (payroll, accounting, email), self-hosted or fair source systems for the data-sensitive, and owned/in-house developed systems where the business is unique. The art isn't choosing one model for everything, it's choosing deliberately per system instead of letting random trial accounts decide the architecture.
The GDPR and exit perspective: can you get out again?
Data protection and exit are more closely connected than most software decisions reflect.
The responsibility is yours, whoever hosts
GDPR applies whether the software is rented or owned. But the roles are different: with SaaS the supplier is typically the data processor, and you need data processing agreements, a handle on sub-processors and an overview of which countries data is processed in. The EU's common GDPR rules make companies responsible for how personal data is collected, stored and passed on, including the duty to be able to hand over or transfer a person's data, and that responsibility can't be outsourced along with the operation. If you own or self-host the system, the chain is shorter: fewer parties, fewer agreements, more direct control, but also full responsibility for security, with no supplier to lean on.
For companies that use AI in customer-facing processes, there is also an extra layer of regulation on top, which we have covered separately in our guide to the EU AI Act for Danish sales organisations.
Data portability: a right for individuals, and an exercise for you
Under GDPR, data subjects have the right to data portability, to have their information handed over in a structured, machine-readable format and moved to another controller. The Danish Data Protection Agency's guidance material on data subjects' rights describes what the right means in practice. For the software choice the point is this: if your system can't export data properly, you can neither fulfil that kind of request elegantly, nor get out of the system yourselves the day you want to switch. Test it before you sign: ask for a full export during the trial period, and look at what you actually get. Raw CSV files without relationships are not an exit option; they are a hostage situation with nicer formatting.
The exit plan belongs in the decision, not in the crisis
A software decision without an exit plan is half a decision. Specifically, before the choice is made you should be able to answer: What can be exported, in what format, and how often? What does it cost (some suppliers charge for export or API access at scale)? How long would a migration realistically take? And what happens to data after termination, are they deleted, and can you get documentation of it? In the ownership model the answers are trivial. In the rental model they are a negotiation, and the negotiation is easiest while the supplier is still trying to win you.
Decision framework: the questions that decide the choice
Here are the dimensions the choice really rests on, and the questions you should ask per system, not once for the whole company.
Owning vs. renting: qualitative comparison on the most important decision dimensions
| Dimension | Renting (SaaS) | Owning / self-host |
|---|---|---|
| Time to get started | Very short (account the same day) | Longer (setup or development first) |
| Cost profile | Low start, grows with users and tiers, supplier-controlled increases | Higher start, flat operation, self-controlled prioritisation |
| Data control | Data with the supplier, access via export/API | Data in your own database under your own control |
| Customisation | Within the supplier's framework and roadmap | Free, limited by your own capacity |
| Maintenance and security | The supplier's responsibility (invisible to you) | Your responsibility (requires a plan and discipline) |
| Exit | Depends on export options and agreement (test before buying) | Trivial (you're already there) |
| Dependence on the supplier's fate | High (closure, acquisition or change of course hits you directly) | Low, especially with open source / fair source code |
The seven questions you should ask per system
- Is the system close to the core? The closer it is to what makes you who you are (your customer data, your process, your pricing), the stronger the case for owning or self-hosting. Standard tasks far from the core (payroll, mail) are rented with a clear conscience.
- How many will need to use it in three years? Per-user prices make growth expensive. A system the whole organisation has to join argues for owning or for licence models without per-user payment.
- How unique is your process? If the standard template fits, renting is fine. If the system has to be bent a lot, you pay twice in the rental model: first for the software, then for the work of getting round it.
- What is the exit cost, tested, not assumed? Do a trial export before you sign. If the answer is unpleasant, it's better to know now.
- Who takes operational responsibility, if you own? Owning without a named person responsible for updates, backup and security isn't a strategy, it's postponed technical debt.
- What happens if the supplier disappears or changes course? For SaaS: what does the agreement say? For fair source/open source: the code is the answer.
- Is there a middle way? Always ask whether the system exists in a self-host or fair source variant before you choose between the extremes.
Note that the questions aren't about technology, but about business. That's also why the decision shouldn't sit with whoever happened to set up the trial account, it belongs in the same place as other long-term agreements on finances and risk. How software choices interact with the broader commercial strategy is something we have written more about in our guide to SaaS sales strategy, which by its nature sees the matter from the seller's side, something that is instructive in itself for any buyer.
Typical pitfalls with both routes
Both models have well-known ways of going wrong. Know them before you choose.
Pitfalls of renting
- The sticker-price decision. Choosing on monthly price per user without factoring in growth, tier creep and exit. The real comparison is the total cost over the period you realistically stay in the system, and that period is almost always longer than you think, precisely because of lock-in.
- Trial-account architecture. The company's system landscape is shaped by random free trials that became habits, instead of by decisions. The result is overlapping subscriptions, data spread across many suppliers and no overall owner of the landscape.
- The integration glue. Every rented system is an island, and the glue between the islands (integrations, exports, double entry) becomes an invisible, growing cost no invoice shows.
- Exit postponed until the crisis. Export options are only investigated the day the price has gone up or the supplier has been acquired, when the negotiating position is at its weakest.
Pitfalls of owning
- Underestimating operations. The build is the visible part; operations, security and updates are the long tail. Without a plan for the tail, owned software turns into technical debt with your name on it.
- The Torben problem. The system only one person understands. AI agents ease the problem (code can be explained, documented and handed on far more easily than before) but the discipline of documentation and shared knowledge is still your responsibility.
- Building what can be rented cheaply. Owning makes sense close to the core. Developing your own payroll system is almost always a waste, the standard product is better and cheaper. Build the unique, rent the generic.
- The everything-at-once project. The big system that is to replace five rented systems in one go fails more often than the gradual route: one system at a time, in operation and proven, before the next is taken on.
- Owning without exit thinking the other way. Owned systems, too, must be able to be wound down or handed over. Your own code without documentation and without version control is its own form of lock-in, just to the past instead of to a supplier.
Conclusion: choose deliberately, system by system
Owning or renting isn't an ideological choice, but a portfolio choice. Rent the generic, where the convenience of the SaaS model is the whole point. Own or self-host what carries your data and your distinctiveness, and use the middle forms (self-host, fair source) where they exist, because they give control without requiring you to build everything yourselves. What's new in the sum is that AI agents have made the ownership side realistic for companies that have never had a development department, and that makes it worth taking the decision again for the systems where the rental model has quietly become expensive, rigid or both over the years.
Whatever you choose: make the choice with open eyes, test the exit route before the entrance, and don't let a trial account from three years ago be your software strategy.
Frequently asked questions
Q: Isn't renting software always the cheapest?
In the short term, yes, the starting cost of SaaS is low. But the rental price is coupled to your growth (per user), to the supplier's tier logic and to price increases you don't control. Over a period of years, and especially for systems the whole organisation uses, the ownership or self-host model can be cheaper overall. The right answer depends on the type of system: calculate the total cost over the period you realistically stay in the system, not on this month's sticker price.
Q: Can a company without an IT department realistically own software?
Increasingly, yes. AI-assisted development has moved the bottleneck from "can we write the code?" to "can we describe the need precisely and take operational responsibility?". The building work can today to a large extent be done by AI agents under human direction. But operational responsibility (updates, backup, security) doesn't disappear; it has to be placed with a named person or a partner before owning is defensible.
Q: What is fair source, and how does it differ from open source?
Fair source is a licence model in which the supplier runs the product commercially, but the code is available and after a defined period converts to genuine open source. The difference from classic open source is the time delay; the difference from classic SaaS is that the code isn't secret, and that customers have a concrete guarantee in the licence if the supplier closes or changes course.
Q: How do I test a SaaS supplier's exit options before I sign?
Ask for a full data export during the trial period and assess what you actually get: Are the relationships between the data preserved, or are they flat files? Is the format structured and machine-readable? Does export or API access cost extra? Also ask in writing what happens to your data after termination, and whether deletion is documented. The answers belong in the basis for the decision, not in a future crisis.
Q: What does GDPR require when we rent software rather than own it?
The responsibility for personal data is yours in both models. With renting, the supplier is typically the data processor, so you need a data processing agreement, an overview of sub-processors and of where data is processed. With owning/self-host the chain is shorter and the control more direct, but the full security responsibility is yours. Data subjects also have the right to data portability (to have their data handed over in a structured, machine-readable format), which in practice requires your system to be able to export properly, whatever the ownership model.
Q: Should we move all our systems to the ownership model at once?
No: it's a classic pitfall. The gradual route works best: pick one system where the rental model hurts most (typically where the per-user price, the rigidity or the data control is the problem), move or build it, get it into stable operation, and then take the next. Generic systems such as payroll and accounting can with advantage stay in the rental model; it is the systems close to your core business that are candidates for owning.
#software-strategy#saas#self-host#fair source#data-lock-in#gdpr#ai-agents#decision-guide
Contents
Unsure about the rules for your own set-up?
We look at what you use today and what it requires: documentation, consent and what the EU AI Act means for you. 20 minutes, no commitment.
Book a reviewRelated articles

Claude Code in a business: what an AI developer agent can do, and what it requires
What Claude Code, Anthropic's agentic developer tool, can do in a business, and what it requires in security, governance and organisation.

AI strategy for businesses: a step-by-step guide
A practical guide to an AI strategy: map workflows, choose use cases, fix your data, run a pilot, bring staff along and know when not to automate.

AI Barriers in Danish SMEs: What Is Holding Businesses Back?
In an autumn 2025 Danish SME survey, 44% had worked with AI and 10% had implemented it systematically. Five barrier themes and possible first steps.