AI agents for business that do the work, with you in control
An AI agent does a job with several steps. It reads data, uses your tools, makes a draft, checks it and hands it on. I build custom AI agents for companies without a developer of their own, and for teams that want someone who has run agents in production.
I'm all in on AI. I use my own agent workflows, so I move fast. I stay careful about what an agent is allowed to change and how you can check its work.
I usually reply within one working day.
What AI agent development covers
Agents that do a task in your business, with limits and a record of what they did.
Custom AI agents
An agent for one job in your business, like research, drafting, sorting requests or preparing reports. It works with your data and your tools.
Agents that use your tools
Agents that read from and write to your CRM, your database, your shop or your internal apps through their APIs. Each agent only gets the access it needs.
Several agents on one job
Bigger jobs split into steps, with one agent per step. The steps run in queues, and every result is saved, so a failed step can run again.
Checks and approvals
Structured output, validation and a human approval before anything important goes out, like an email to a customer or a change to your data.
Cost and provider choice
The right model for each step, limits per job, and a clear view of what a run costs. I work with Anthropic, OpenAI, Google and xAI models, and you can switch providers later.
Logs you can read
Every run leaves a record of what the agent saw, what it did and what it cost. When something goes wrong, you can see why.
How I work
You don't need to have everything figured out before we talk. An idea, access to the repo, a screen recording or a list of problems is enough. I look at what is there, tell you what I would do and what it costs, and then I build it.
I use Claude Code, Codex and my own agent workflows for reading code, planning changes, writing tests, refactoring and debugging. It is not vibe coding. It is senior engineering with a much faster loop.
AI-first, and I run agents myself
My own work runs on agent workflows with Claude Code, Codex and tools I built. So I know where agents go wrong. Every agent I build has limits, approvals and logs.
I stay in control
AI writes a lot of my code. I still make every technical decision, read every change before it ships, and test it. Fast, but meticulous: the speed is only worth something if what goes live is right.
The way you want to work
Some clients want to look at every change before it goes live, others want me to deploy. Some want a weekly call, most prefer written updates. I work well asynchronously, and I still ask the important questions when they come up.
I think like an owner
I still run my own online products. So when I build or fix yours, I think about traffic, conversion, revenue and support as much as about the code, and I tell you when something is not worth building.
Ways to work with me
Fixed price, a flat monthly rate, or hourly. It's always me doing the work, not an agency. If you're not sure, describe what you need and I'll suggest one.
Examples of my work
Agents and automated workflows I have built.
Many AI agents per job, across four providers
A Laravel and Vue product that orchestrates many AI agents on queues, with Anthropic, OpenAI, Google and xAI models. The agents use scraped data from search engines and publish to WordPress through its API.
Runs on its own across 30+ servers
A Laravel platform that finds link opportunities, discovers contact details and sends personalised email campaigns. It processes millions of opportunities a month without anyone running it by hand.
My own AI software delivery product
vroni is my own product for AI software delivery. It runs on agents, so I see their mistakes and limits up close.
What clients say
"He doesn't need micromanagement - he's fully capable of making decisions and driving the project forward, which has freed me up to focus on other parts of the business."
"Vincent does not just close tickets, he thinks through the product, the edge cases, and the customer impact."
"He then built exactly what was needed and kept us informed throughout. Communication was clear and proactive at all times."
Questions
What is the difference between an AI agent and a chatbot?
A chatbot answers questions. An agent does a job with several steps, uses your tools and produces a result, like a report, a draft or an updated record.
Can an AI agent work with our existing software?
Yes, if the software has an API or a database I can reach. The agent gets its own access with only the rights it needs.
What happens when the agent makes a mistake?
It will sometimes. I build validation and an approval step where a mistake would cost money or trust, and every run is logged so you can see what happened.
Do you build agents on n8n or in code?
Both. Simple flows can live in a tool like n8n. When the agent needs your data, limits and logs, I write it as code inside your app or as a small service.
Which AI models do you use?
Whatever fits the step. I have built with Anthropic, OpenAI, Google and xAI models, and cheaper models do the simple steps.
How do you charge?
Fixed price for a job with a clear outcome, a flat monthly rate for ongoing work, or hourly. Describe what you need and I'll suggest one.
Get a quote
Which job you want the agent to do, which tools and data it would use, and who checks the result. A few sentences are enough. I'll reply with how I would build it and what it would cost.
If you would rather talk first, book a 30-minute call or email contact@vincentschmalbach.com.