// ai_consultant · jönköping
An AI consultant in Jönköping that builds, not just advises
Search for an AI consultant in Jönköping and you mostly find agencies selling workshops and reports. We do something else. We are an engineering shop that writes code. You get a system that works, not a slide deck about what AI might do one day.
Jönköping and Småland mean manufacturing, logistics, retail, and family-owned companies that have built something over decades. That is where the real data sits: order history, drawings, support email, quotes, stock levels. That data can be useful today, if someone wires it to a language model the right way.
We are based in Jönköping and happy to come out to your factory or office. We also work remotely with companies across Sweden. If AI is the wrong tool for your problem, we say so on the first call.
What an AI consultant should actually do for a Småland company
An AI consultant should not sell you a strategy. The consultant should find the boring job that eats time, build something that removes it, and leave you with a system your own staff can use and understand.
Here is what that looks like in practice. We walk through how you work today: where the email lands, where quotes get written, where someone types the same numbers in twice. Then we point at one or two places where a language model does clear good. Not ten. One or two that can be built and measured.
We would rather turn a project down than build something nobody will use. Many AI projects die because they started at the wrong end: with the technology instead of with the job to be done. A machine shop in Vaggeryd does not need an AI strategy. It needs to get rid of the three hours a day spent answering the same questions about delivery times.
Typical first projects
This is the kind of thing we build. No case studies with made-up numbers, just concrete jobs that fit companies your size:
Support agent
An agent that reads incoming email and tickets, answers the common questions about deliveries, spare parts and invoices, and hands the hard ones to a person with the history already pulled up.
Internal search over documents
Drawings, manuals, contracts, old quotes and email threads in one searchable layer. Staff ask a question in plain language and get the answer with the source next to it, instead of digging through folders on the file server.
Quoting and offer automation
An agent that reads an inquiry, pulls the right items and prices from your ERP, and drafts a quote the salesperson reviews and sends. Hours become minutes.
Forecasting from ERP data
Order history, seasonality and lead times from Monitor, Jeeves, Visma or whatever you run, turned into a plain forecast for purchasing and staffing. Nothing magic, just your own data put to use.
How an engagement runs
Step one is a twenty-minute call. You tell us what takes time, we ask questions. It costs nothing and you commit to nothing.
Step two is a scoped, fixed-price job. You know what it costs before we start. No hourly meter running, no surprises on the invoice. If the job is too big to price honestly, we split it into smaller pieces.
Step three is working software in weeks, not quarters. We build the first version, test it against your real email and documents, and put it in the hands of your staff. Then we adjust based on what actually happens.
No lock-in. The code is yours. We build on Claude from Anthropic and on open-weights models you can run yourself. If you want to take over operations or switch vendor tomorrow, you can. We want you to stay because it works, not because you are stuck.
Where the data lives
Your data stays in the EU. We operate servers in Germany and Finland and use Swedish hosting when it is needed. Nothing goes to a US cloud and nobody trains a model on your customer data.
GDPR is not a box we tick at the end. We build so that personal data is used only for the job, logged in a way that can be audited, and deleted when it is no longer needed. If you have a data protection officer or an IT manager who wants to see how it fits together, we are happy to show them.
If the data must never leave the building at all, we run the model on a server at your site. That fits when drawings or customer data are sensitive enough that an external server is the wrong answer.
Why local matters
An AI project goes better when the consultant has stood on the shop floor. We can sit in your factory for a day and watch how an order actually moves from email to delivery. That does not show up in a Teams meeting.
We know the region. We know that many companies in Gnosjö, Värnamo, Nässjö and Jönköping run older ERP systems, have an IT department of one or two people, and want something that works without depending on a big-city agency. That is exactly the situation we build for.
And when we do not need to be on site, we are not. Most of the build happens remotely, and you see progress as it happens. The combination is the point: local presence when it makes a difference, remote when it is more efficient.
// faq
Frequently asked questions
What does an AI consultant cost?
We do not bill by the hour. We scope a defined job and give a fixed price before we start, so you know what it costs. The first call is free. The price depends on how many systems need connecting and how much data goes in, so we do not guess until we have seen your case.
Do we need our own data?
You already have more data than you think. Email, quotes, order history, manuals and support tickets go a long way. We do not need a prepared dataset or a data warehouse. We start with what exists and clean what needs cleaning along the way.
How fast can we start?
We can have the call this week. A scoped project usually starts within a few weeks depending on our workload. A first working version of a support agent or internal search typically takes a few weeks to build. Larger ERP integrations take longer, and we say so up front.
Do you only work in Jönköping?
No. We are based in Jönköping and happy to visit companies across Jönköping County and Småland, but we work remotely with clients across Sweden and the Nordics. Many projects are fully remote from start to finish.
Which AI models do you use?
Claude from Anthropic when the task needs a strong model, and open-weights models like Mistral, Llama and Qwen when the data has to stay on your own server or when a smaller model is enough. We do not use OpenAI or GPT.
// read_next
Got a job that eats time every day?
Tell us what it is. We take a twenty-minute call, say honestly whether AI is the right tool, and give a fixed price if it is. We are happy to come out to you in Jönköping.