AI development agency Rotterdam

An AI development agency builds software in which language models and automation work together with the systems an organisation already uses. Think of an assistant that answers from your own documents, a process that assesses applications automatically, or an integration that brings data from four systems together. A small part of that work is calling a model, and the largest part is everything around it: opening up the data, checking the output, catching the failure and keeping the whole thing running. An agency that only advises stops at the design, while a development agency hands over the working system.

Usually at the moment an idea outgrows the test setup. A prompt in ChatGPT that works well for one person is something else than an application ten colleagues lean on daily and where a mistake has consequences. As soon as you want to point AI at your own data, or connect it to a CRM, ERP or invoicing system, you run into questions about permissions, cost and reliability that someone has to decide on. Once a model's answer lands somewhere in a process people rely on, you are building software.

A first working application usually takes us a few weeks of work, and climbs from there as more systems hang off it and the cost of a mistake rises. The model itself is almost always the smallest line: a few tens to a few hundred euro per month in API costs, even under serious use. The money goes into opening up your data, catching the cases where the model gets it wrong, and the maintenance afterwards. The question you want to put to an agency is what it costs to keep the application running reliably for a year.

Because a demo and a production system solve different problems. In a demo you ask the question the model is good at, and in production people ask the question you did not foresee. What comes with that is access to current data, a way to see when an answer was wrong, costs that move with usage, and someone who is responsible when it goes wrong. That layer is missing from most pilots, and that layer is the actual work. So we start from the question of what should happen when the model gets it wrong, and build backwards from there.

An AI consultant thinks along about where AI delivers value and hands that over as advice, a roadmap or a business case. An AI development agency builds the system and carries the responsibility that it keeps working. In practice those roles run into each other, because most choices about where AI makes sense only get sharp once you know what it costs to build them. We do both from the same people, so the advice comes from someone who has to make it true afterwards. We build software; AI campaigns and content marketing we leave to agencies that are good at that.

AI sits on both sides of the work here: we build applications with it for clients, and our own process runs entirely on it, with a human checking every line before it goes live. That means we know from experience where generated code and generated answers go wrong, because we run into it daily. We push back when AI is the wrong route, also when that means less work for us. We are a senior studio at Hofplein in Rotterdam, so you can drop by and watch what is being built at the table.