Computer vision, RAG pipelines and LLM integrations from €8.000. Claude, GPT and local models. Production AI that runs for months.
Custom AI development at NedDev starts at around €8,000 for a focused application and rises as the integrations and the scale increase. We build AI that runs in production: RAG knowledge bases, document extraction, voice agents and LLM integrations on a Laravel or FastAPI backend, with logging, fallbacks and cost monitoring built in. A first working version on real data is usually live in 6 to 10 weeks. We do not deliver a demo that stays stuck on a laptop. NedDev is a The Hague studio, KvK 65641922.
Custom AI development means you have a software system built that carries out tasks that previously needed a person: reading documents, answering questions based on your own knowledge, holding conversations or pulling data out of forms. The difference with a toy is that a real AI solution becomes part of a work process and runs on it day in, day out.
The price depends on the complexity and the integrations. An AI assistant that gives answers based on your own documents is cheaper than a voice agent with a telephony integration or a system that reads drawings and calculates quantities. Alongside the build you have ongoing costs for the use of the AI models themselves, which we make clear up front and monitor with hard limits.
Our price ranges for custom AI development:
Most AI projects do not founder on the model, but on everything around it. A demo works in a controlled situation. Production means dealing with messy input, high load, costs that can run up and users who use the system in a way nobody foresaw. That is where the real work is, and that is where a working system sets itself apart from an impressive presentation.
So we build with production in mind from the first week. That means logging on every AI call, fixed limits on costs and a fallback when the model is uncertain. For our AI colleague Cor we built a platform with RAG memory that stays separated per client. For Lexi AI we delivered an assistant that answers collective labor agreement questions based on recorded sources, citing the article the answer rests on. For JinSulate a geometric counter replaced the loose AI estimates, so the same drawing always gives the same result.
Does AI work even if I have no technical team? Yes. We deliver the system including management, monitoring and a dashboard where you see for yourself what is happening. You do not have to hire a prompt engineer to get value out of it. We make sure it runs, you use it.
What we build in as standard with custom AI development:
The biggest fear with AI is that it confidently sells nonsense. That risk is real if you deploy a model on its own without controls around it. We reduce it with RAG, where the model only answers based on recorded sources instead of guessing from memory. The answer cites the source, so a user can check it instead of trusting blindly.
For numerical tasks we go a step further: there we replace the model's guess with a fixed calculation. At CaseMeister and FlexUren that meant reliable outcomes instead of varying AI estimates. The step-by-step plan we follow on an AI project:
Which AI model do you use? We are model-independent. We choose per task the model with the best ratio between quality and cost, and we build so you can switch later without rewriting the whole application. For sensitive data we advise, where needed, models that run within the EU.
Responsible AI use is not a side issue. The Dutch government publishes practical guidelines on this, see rijksoverheid.nl. We take that into account, certainly for applications involving personal data or decision-making.
Our AI runs on the same stack as the rest of our work, so it stays manageable. The backend is usually Laravel or FastAPI, with a separate layer for the AI calls. For platforms that serve multiple clients we connect this to our multi-tenant SaaS architecture, so client data never gets mixed up and each client keeps its own walled-off knowledge base.
What AI can do for a business in practice:
The hallmark of a good AI solution is predictability: you know what it costs per month, you see what the system does, and you can read back why it gave a particular answer. Want to know whether your process lends itself to AI? Take a look at our approach to AI development and how we combine it with SaaS platforms.
Our internal stack packages for multi-tenant SaaS. A Laravel + Filament starter, an audit-trail engine, and a tenant-impersonation package that runs across 12 clients.
8 PACKAGES · 2.4K STARS EDITORIAL · LONG-FORM ↗What we write down as we learn it. Case studies, technical write-ups, design decisions. No content marketing, just real knowledge.
42 ESSAYS · MONTHLY RESEARCH · AI EXPERIMENTS ↗Side projects and R&D. Voice-agent prototypes, RAG pipelines, AI knowledge-graph experiments. Some become products. The rest teach us something.
14 EXPERIMENTS · LIVE DEMOSCustom AI development with us starts at around €8,000 for a focused application, such as a RAG assistant on your own documents. AI with integrations and an admin environment usually costs between €15,000 and €40,000. An AI platform with voice, document extraction or computer vision starts around €40,000 and grows with the scale. Alongside the build you have ongoing costs for the use of the AI models themselves, plus hosting and management. We make those monthly costs clear up front and monitor them with hard limits, so a spike in usage never causes a surprise on the bill. We would rather give an honest range after a short conversation than a number that looks too good up front.
A first working version on real data is usually live within 6 to 10 weeks. We work in short sprints with weekly demos, so you see the system grow and can steer before it is finished. The first weeks after launch we follow along on real input, because that is the moment when it becomes clear whether the system also holds up outside the test environment. A large, organization-wide AI platform takes more time of course, but we cut that up into parts that deliver value on their own, so you do not have to wait months for a first result. That way you see early whether the approach works before the whole budget is spent.
That risk exists if you deploy a model on its own without controls around it. We reduce it by working with RAG, where the model only answers based on recorded sources instead of guessing from memory. The answer cites the source, so a user can check it. On top of that we build in a fallback for when the model is uncertain, and we log every answer so you can see back why something was said. For numerical tasks we replace the model's guess with a fixed calculation, as we did at JinSulate. That way the result becomes predictable and verifiable instead of a guess that happens to turn out right or wrong.
Yes, after delivery the source code is yours, including the admin environment and the documentation. You are not tied to us. What is different with AI than with ordinary software: the AI models themselves usually run at an external provider, and using them costs money per month. We build the system so you can switch models without rewriting the whole application, so you are not stuck with a single supplier. The knowledge base, the logic and the integrations are yours. If you want another team to do the ongoing development later, that is possible with the documented code we deliver.
The use of the AI models is billed per use, usually per amount of text processed, and those costs run through your account with the model provider or through us if you prefer to outsource that. The amount depends on how much the system is used and which model we choose. For an assistant with a few hundred questions a month it is modest, for a heavily used platform it runs up. That is why we build in cost monitoring as standard, with hard limits per user and per month, so a spike never causes an unexpected bill. We make these monthly costs clear up front, so you know where you stand before the system goes live.
