SökHem
Buyers describe what they want once, draw their areas on a map and get matched every day when the right home is listed. Sellers list for free, with or without an agent.
- 130,000+
- homes for sale
- 2.5 M
- matches delivered
- 750+
- area guides
AI development for companies
Custom AI solutions in your stack. Once you know where AI pays off, the next step is to build it for real. I develop agents, automations and integrations that take over contained steps of your work, in your stack and to the same quality bar as the rest of your systems.
Book an intro callWhat I build
Most of what I build moves routine work from people to systems, without the people losing control.
Every solution starts narrow. One agent that fully handles one step beats one that half handles ten.
- Agents that read, sort and answer incoming email and tickets based on your own help articles and routines.
- Automated finance flows: matching invoices, chasing receipts and suggesting how transactions should be coded.
- Agents that summarise, translate and compare long documents and flag what needs a person’s judgement.
- Reports built automatically from numbers in several systems, with a summary of what has changed.
- Integrations with your CRM, business system and document stores through APIs and MCP, so agents work in the tools you already have.
Built to last
An agent that works in a demo is easy to build. An agent that works every day for a year takes more. So I build with tests against real examples, logs of every request and tool call, cost tracking per process and clear limits on what the agent may do on its own.
Steps where a mistake is expensive go through human approval. The agent prepares, a person decides, and everything is kept so you can see afterwards who did what. These are the same principles regulated environments require, and I have built software in such environments, including medical technology.
In your stack, not mine
I build in the technology you already have when that makes sense, or in a light, modern stack with TypeScript, Node or Bun and Postgres when you start from scratch. Models are chosen per task: sometimes a large model from Anthropic or OpenAI, sometimes a smaller open model that runs on your own server.
The code is yours from day one. It lives in your repository, is documented, and can be taken over by your team or another supplier whenever you like.
How it runs
- We start from the assessment, or a short walk-through of the process if you already know what to build.
- I build a first version against real examples from your business and measure how often the agent gets it right.
- A few people on the team use it in their normal work, and we adjust until it meets the bar.
- It goes into production with monitoring, and I run it for you or hand it over to your team.
A good fit when
- You know which process to automate but have nobody to build it.
- You have a pilot that worked in a demo but never reached production.
- Your systems need to talk to each other before AI can help.
- You work in an industry where traceability and approvals are required.
A typical first engagement
A typical first build is one agent for one process, such as incoming customer email. It can usually be running within a few weeks, and then we build on what it actually saves.
Read more about how I work as an AI consultant in Gothenburg.
Related services
Examples
Wez
My own agent harness, where a crew of AI agents works in my projects. Every request, tool call, file change, cost and approval is visible while it happens. It is how I build, and you get the same visibility in what I build for you.
- 1,300+
- commits
- 6
- agents in the crew
- 3
- clients: desktop, mobile and CLI
Read the casePrivate project, in daily use
Questions
Which AI models do you use?
Whichever fits the task best in quality, cost and where your data may be processed. I use models from Anthropic, OpenAI and others, and open models that can run on your own server, and I build so the model can be swapped.
What happens when the agent gets something wrong?
Every agent has clear limits on what it may do alone. Risky steps need human approval, everything is logged, and we measure accuracy against real examples before it gets more responsibility.
Can you build in our existing technology?
Yes, in most cases. I have built in everything from large enterprise environments to new platforms, and I adapt to your stack, your code review routines and your security requirements.