DeepL and the big question: What can an ERP really do?

ERP systems are often considered a necessary evil. But DeepL shows that they can achieve much more strategically – if you don't just digitise them, but rethink them. A story about growth, data intelligence and operational excellence in finance.

Blog header for blog post: DeepL and the big question: What can an ERP really do?

There is a certain kind of silence that spreads through organisations when the conversation turns to ERP systems. A silence that lies somewhere between latent frustration and premature resignation. Anyone who has worked in a growth-oriented company over the last ten years knows this feeling: Your own financial system seems like a relic from another era. Processes are slow, data is fragmented, responsibilities are blurred. And somewhere in all this, a paradoxical situation arises: the faster a company grows, the more it loses business transparency. DeepL, the company behind one of the world's best-known AI-based translation platforms, was at exactly this point. The product was booming.

New markets, new user groups, a growing B2B customer base. At the same time, financial processes were stalling. External accounting, isolated solutions, lack of scalability. CFO Marcus Hada described the situation with a candour that is rare in this industry: ‘We had to build the ship while we were already sailing it.’

But how do you react when the system in the background can no longer keep up with the pace of the organisation? When momentum becomes a burden?

Why finance teams often struggle more with Excel than with strategy

In public, companies like to talk about innovation, AI and strategic transformation. Internally, however, many struggle with completely different problems: payments that cannot be correctly allocated. Invoices that have to be matched by hand. Reconciliations that take days to complete. In theory, we live in the cloud – in practice, we live in spreadsheet chaos.

The real change only becomes apparent after the go-live.

One of the key challenges lies in the lack of connection between data systems. Cash flow information is isolated from customer master data, and accounting systems are not integrated with payment systems. When it's time for the monthly closing, the financial puzzle starts all over again.

In the midst of its growth, DeepL decided to revamp its system landscape. Not just a few isolated improvements, but a whole new infrastructure. The goal was to achieve a common understanding of data, greater automation and a clearly defined framework for the work of the finance department.

What happens when you don't tinker with the system, but change it

Enterprise software is rarely attractive. But its impact is all the more far-reaching.

Choosing a system like Workday changes the technical landscape and also the way processes, responsibilities and standards are discussed within the company. DeepL took on this challenge. With a small team and limited in-house experience, they began the implementation. The focus was on creating a sustainable system landscape for the future – by consciously moving away from mere digitalisation towards a targeted structural realignment.

The implementation took eight months and was both a learning process and structural work. The decisive factor was the openness to embrace proven standards and question familiar patterns.

However, the real change only became apparent after the go-live.

How DeepL transformed its ERP into a learning system

Andrzej Szymanski, responsible for the financial system landscapes at DeepL, approaches the topic through thought processes: he shows how the right tools can be used to analyse and structure operational problems at a deeper level.

A practical example: a large number of bank transfers from corporate customers were received, often without any recognisable references to invoice numbers. In smaller structures, such cases can still be checked manually. But with thousands of transactions, this becomes a real challenge. Szymanski developed a matching logic that works within the system like a neural network without AI – based on experience, data structure and comprehensible rules. The results spoke for themselves: over 85 per cent of payments were automatically allocated. The proportion continues to grow.

Systems must keep pace – or they slow things down.

In B2C business, the situation was even more complex: more than three million individual transactions per year, processed via external payment providers. The solution was to first consolidate all data via Workday Prism and then only highlight the discrepancies. In the end, only 15,000 individual cases had to be checked manually.

How systems become the location of business intelligence

The remarkable insight from this story lies less in the technology than in the attitude with which it was used. At DeepL, the ERP system was seen as a creative space for better processes.

This view is unusual. In many companies, ERP systems are considered background technology. But they could do more: provide a framework for operational intelligence, for the targeted interaction of data, for the continuous improvement of processes.

The origin of many problems lies in the history of these systems. They were developed for times when processes were stable, data was slow and organisations were clearly structured. Today, however, conditions are different: agility, real-time capability, network structures. The systems have to keep up – or they slow things down.

Without a solid data structure, AI remains just a promise

The discussion about artificial intelligence often neglects a fundamental fact: without structured data, even the most advanced application is ineffective. Systems such as Workday provide the basis for this. They enable a consistent model for central business functions and uniform configuration instead of rigid programming.

How much these opportunities are exploited depends on the company itself. DeepL has turned this into a strength. The way data is handled there demonstrates a new understanding of roles in the finance department. It is no longer just a matter of checking figures – but of using them to prepare better decisions.

The future is created where systems are part of an adaptive, creative organisation. ERP systems are not problem solvers, but they do reveal weaknesses by distinguishing successful processes from inefficient organisational structures. DeepL shows how technical infrastructure can be used strategically and that progress depends less on new tools than on the effective use of existing tools. An adaptive and creative organisation that uses its systems consciously is the key to future success.

More Reading