What does automating business processes actually cover?
It is an umbrella term that covers many disciplines in practice: workflow automation (quotes, invoices, complaints), document automation (contracts, forms, incoming mail), RPA (bots in legacy systems without an API), process orchestration via BPM platforms with BPMN and DMN models, SaaS integrations via iPaaS, AI-augmented automation (LLM classification, document AI), data pipelines (ETL and event streaming) and setting up approval and escalation routes. A mature project almost always combines several of these categories — pure RPA without integrations is no longer the norm in 2026, and iPaaS-only is too limited for most mid-market organisations.
What is the difference between RPA, BPM and iPaaS?
RPA (UiPath, Blue Prism, Automation Anywhere) mimics human user actions in applications without an API, making it a temporary bridge to systems that do not support integration. BPM (Camunda, Flowable) is a process engine that tracks the status of long-running processes, assigns tasks to people, and orchestrates approvals, escalations and deadlines. iPaaS (Workato, Make, Zapier, Power Automate) connects SaaS applications using ready-made connectors and is best suited to relatively simple event-driven flows. They are not mutually exclusive; in a mature process landscape we see them side by side for different types of flows, often with a BPM engine as the orchestrator that calls iPaaS and RPA bots.
When do you choose low-code and when custom?
Low-code (Power Automate, OutSystems, Mendix) suits linear flows with standard building blocks, a limited number of exceptions and a reasonable volume. Custom development pays off when a process runs through many systems, is subject to industry-specific regulations, needs to be multi-tenant for a cooperative or trade organisation, or has AI components intertwined with human decisions. A second signal: as soon as you spend more than forty per cent of your time in low-code building workarounds for its limitations, you lose the cost advantage and custom development can work out cheaper. When in doubt, we often start with low-code for the first tier and migrate to custom development if the complexity eventually demands it — that is almost always a healthier order than the other way round.
What determines the cost of an automation project?
The biggest cost drivers are: the number of systems to integrate and the quality of their APIs, the number of exceptions and the extent of sector-specific regulation, the number of human approval steps, the requirements for auditability and compliance (GDPR, DORA and NEN 7510 each have their own impact), the choice between platform licences and a custom build, and the extent to which your own team takes over management after go-live. We work with fixed sprint budgets so you can steer scope and investment per iteration, and at the end of each sprint we give you an honest estimate of the expected follow-on costs — no open-ended commitments, no surprises afterwards.
How long before we see results?
A first automated process usually goes into production after a handful of sprints. A mature process landscape with multiple flows, monitoring and ongoing development takes a programme of several sprints and continues afterwards as an ongoing programme. We deliberately avoid hard week counts — these depend heavily on scope, the number of systems involved, your team's availability and the degree of organisational change the process brings with it. What we do promise: after the first sprint you will see the first working automation or a tangible partial flow, not just documentation.
How do we calculate the ROI of process automation?
In the analysis phase we measure, per process: manual time per transaction, monthly transaction volume, error rate and recovery costs, lead time and the cost of waiting times (think late invoices, delayed quotes). This gives us a baseline. After automation we compare the same metrics over a comparable period. Additional value — shorter lead times, a better customer experience, compliance assurance, scalability without extra staff — we quantify where possible, but it is harder to express in euros. We present the business case with supporting evidence rather than as a marketing promise, so you can decide for yourself where the next investment will pay off.
What is the role of AI in process automation?
AI is used mainly in three places: classifying incoming documents (invoice, complaint, contract, permit), extracting structured data from unstructured input such as emails, scans and PDFs, and suggesting follow-up actions to human reviewers based on historical data. We never use AI fully autonomously for decisions with a major impact — there is always a human in the loop, especially for financial, medical or legal decisions. For every AI component we also build in the logging and audit trail that the AI Act requires for risk-category classification and transparency.
Do you work with our internal IT department?
Almost always. We transfer knowledge throughout the entire engagement, not just at the end. We deliver BPMN diagrams, DMN decision tables, API contracts and runbooks so your own team can manage and further develop the processes. For ongoing management we offer an optional SLA contract with four response-time tiers, but we prefer your team to handle primary management, with us available for escalations and larger changes. That way you build your own capability instead of dependency.
Can you connect to our existing ERP, CRM and HR systems?
As a rule, yes. We have experience with the common ERP suites (SAP, Microsoft Dynamics, Exact, AFAS, Unit4), CRM platforms (Salesforce, HubSpot, Microsoft Dynamics), HR suites (AFAS HR, Visma, Workday) and accounting packages (Exact, Twinfield, Moneybird). For sector-specific systems — think EHRs in healthcare or specialist applications in logistics — we first look at the available API or integration layer, and if there isn't one, we fall back on structured file exchange or, as a last resort, an RPA bot. We don't make "shadow copies" of your data unless that is strictly necessary for performance.