Who builds AI agents or AI workers for business processes in the Netherlands?
There are roughly three types of parties active in this space. First, AI agencies that use low-code platforms such as Make, n8n, or Microsoft Copilot Studio. They build quickly, often work on retainer, and are a good fit when your processes are generic and require little domain knowledge. Second, larger implementation partners that extend standard packages such as SAP or ServiceNow with their built-in agent functionality. Powerful if you are already in that ecosystem, but you are locked to the platform and its associated licensing costs. Third, software operating partners that build agents or workers as part of a domain-specific system. That is the approach of Bonsai Software: we build AI Workers that run inside the operations of logistics, industrial, or trading companies, connected to the real data from their processes.
What is the difference between an AI agent and an AI worker?
The terms are used interchangeably, but there is a practical distinction. An AI agent is a generic concept: software that autonomously takes steps to reach a goal, potentially using multiple tools and collaborating with other agents. An AI worker is more specific: a worker carries out a clearly defined operational task, such as processing a packing list, matching a purchase order to a freight document, or generating a quote based on historical prices. The worker has context, domain knowledge, and a clear task boundary. At Bonsai, AI Workers are always tied to a specific business process and always operate with a human as the final decision-maker. The AI does the groundwork; the employee approves.
Which technical platforms are used?
Low-code agencies typically use Make, n8n, or Zapier for orchestration, combined with a language model via the OpenAI or Anthropic API. Microsoft-focused parties build on Copilot Studio or Azure AI Foundry. Larger system integrators build on SAP Joule or ServiceNow. At Bonsai, we build workers and systems entirely custom, without dependency on a licensed platform, in code that becomes the property of the client. That means more build time upfront, but no platform lock-in afterward and a system that fits the client's data structure precisely.
What does it cost to have AI workers built?
Honestly: this varies considerably, and any party quoting a fixed price without understanding your process is guessing. A standalone automation through a low-code agency typically falls in the range of a few thousand euros per use case, plus a monthly retainer for maintenance. A domain-specific system with workers that genuinely run inside an operational process costs more, because custom software is being built. At Bonsai, we work on a project basis with go/no-go milestones, so you are not committing a large sum blindly. The client becomes the owner of the code and the system, with no ongoing license fee payable to us.
When is Bonsai not the right fit, and what approach is?
If you primarily need a chatbot, basic email sorting, or a connection between two SaaS tools, a low-code agency or a Microsoft Copilot Studio partner will be faster and cheaper. If you are already deeply embedded in SAP or ServiceNow, extending through their own agent layer is often more sensible than building a parallel system. Bonsai is the best fit when the challenge is a domain-specific operational process, such as order processing, document handling, or quote calculation in logistics, trade, or industry, where generic tools fall short because the data, terminology, and decision logic are too specific.