The real cost of "standard"
Until recently, the most important technology decision for a mid-sized company followed a predictable script: evaluate ERPs on the market, pick the closest fit, and spend months — sometimes years — closing the gap between what the software offers and what the business actually needs. In 2026, that script deserves a deep revision.
Implementation costs for major ERPs are well known and, for many companies, hard to swallow. A typical SAP S/4HANA rollout for a mid-sized organisation lands somewhere between €500,000 and several million euros. Microsoft Dynamics 365, Oracle Fusion and similar platforms reach comparable figures once you add licences, consulting, data migration, training and the inevitable annual maintenance.
But the financial cost is only part of the problem. The real toll is rigidity. When a company adopts a monolithic ERP, the business adapts to the software — not the other way around. Processes are redesigned to fit the available modules. Exceptions — those quirks that often constitute a company's competitive edge — are handled with workarounds, parallel spreadsheets or expensive customisations that complicate every future upgrade.
This dynamic has been accepted for decades as a necessary evil. The alternative — building a complete system from scratch — was historically even more expensive, slower and riskier. But that premise no longer holds with the same conviction.
What has changed: AI-assisted development
The emergence of AI-assisted development tools has substantially altered the economics of building software. This is not an incremental improvement — it is a structural shift in the relationship between cost, time and quality.
Development teams that integrate AI into their workflow can automate a significant portion of repetitive code — validations, migrations, tests, API documentation — freeing them to spend most of their time on what truly adds value: modelling the client's specific processes, designing interfaces for real users and building integrations that work the way the business needs, not the way a generic module dictates.
The impact on timelines and budgets is substantial. Projects that three years ago required large teams over many months can now be completed in significantly shorter timeframes and with more contained investments — not because quality is sacrificed (automated test generation and early error detection often raise it) but because an enormous amount of mechanical work that used to consume most of the budget has been eliminated.
There is an irony worth noting: the big vendors themselves acknowledge this shift. SAP is integrating Joule, Microsoft is pushing Copilot, Oracle is embedding predictive models. They invest billions in AI precisely because they know intelligent automation radically transforms how software is built and operated. What they perhaps don't emphasise is that the same AI also empowers smaller teams to build solutions that compete directly with their products.
The maturity of the ecosystem
Another variable that has quietly changed is the infrastructure available for custom software. Modern development frameworks ship with capabilities that a decade ago required months of engineering: robust authentication, permission systems, job queues, caching, audit trails and standard APIs. Cloud infrastructure offers scalability, redundancy and availability levels that were once reserved for large corporations with their own data centres.
In other words, the argument that "only a big ERP guarantees reliability and scalability" has lost much of its technical foundation. The tools to build solid, secure and scalable systems are within reach of any competent team. The barrier is no longer technological — it is domain knowledge and execution capability.
When an off-the-shelf ERP still makes sense
It would be dishonest to argue that custom software is always the answer. There are scenarios where a consolidated ERP delivers value that is hard to replicate: organisations operating across multiple jurisdictions that need certified fiscal, labour and accounting compliance in dozens of countries; sectors with regulations that demand accredited systems; companies with thousands of employees and legacy integration ecosystems where the cost of migration outweighs any potential benefit.
For these realities, platforms from SAP, Oracle or Microsoft will remain relevant. But it is worth asking how many companies currently implementing these systems truly find themselves in that scenario.
The hybrid model as natural territory
Perhaps the most sensible approach in 2026 is not a binary choice. Many companies are adopting a hybrid model: using specialised SaaS tools for commodity functions — accounting, payroll, e-invoicing — and building custom software for whatever constitutes the differential core of the business.
The logic is straightforward: there is no point reinventing payroll calculations or tax model generation. But neither does it make sense to force your sales process, logistics management or customer relationships into a mould designed to satisfy as many companies as possible. The areas where competitive advantage resides are precisely those that benefit most from a system that reflects exactly how the business operates.
With modern APIs as the connective tissue, this model combines the reliability of proven tools with total flexibility where it truly matters.
AI as a product differentiator, not just a process one
There is one final aspect that deserves attention. AI has not only transformed how software is built — it has also transformed what that software can do.
A custom system can incorporate AI models trained specifically on the business's own data and patterns: demand forecasting based on real historical data, document classification adapted to proprietary formats, anomaly detection calibrated with company-specific thresholds. Major ERPs also offer AI features, but their models are necessarily generalist — designed to work acceptably across any sector, not exceptionally in yours.
A decision that can no longer be postponed
Inertia is powerful in corporate technology decisions. Many companies continue to evaluate their next investment cycle with 2018 criteria: find the most complete ERP, negotiate licences, plan a long implementation. And yet the technology landscape of 2026 offers alternatives that were unfeasible just a few years ago.
The question is no longer so much "which ERP do I buy?" as "what technology architecture best serves my business over the next five years?". Answering it honestly requires, at the very least, considering that the answer might not be found in a catalogue.
