AI agent development company
We build and run your AI systems.
No roadmap decks, no advisory retainer. We're developers — we design the system, write the code, and keep it running in production.
Specialised in Claude by Anthropic: we build and run systems on Claude and Claude Code.
What we build
Systems that do the work, not summarize it
Every system we ship does one of six things. Most do two or three at once.
Agents
An agent that reads an incoming message, checks your CRM, and drafts a reply — not a chatbot that hands you a summary and stops there.
Integrations
Built into the tools you already run — Salesforce, Pipedrive, sevDesk and the like. No new system for your team to learn.
Document processing
PDFs, scans, contracts, invoices. Read, structured, and pushed to wherever the data needs to land.
Content and AI-visibility systems
Structured content and product data that search engines and AI assistants can actually read, cite, and recommend.
Outreach systems
Messaging and call layers for contacts who opted in, with every attempt logged against your CRM records — not a spreadsheet someone maintains on the side.
Running it in production
Monitoring, error logs, fixes. A system that fails loud when something breaks, not one that quietly stops working.
We build it, we deploy it, and we're the ones who get paged when it breaks.
The series
Six episodes. Set up your own agent, in order.
We're filming the exact setup we use — from install to an agent that can carry a project for hours.
Install
Desktop Setup
Optional: Your Own Server
Guardrails
Voice Kick-off to PRD
Long-Session Workflow
Customer stories
Built and running in production
Fusion Dental Implants
Patient messaging, appointment confirmations and a call layer, built around their existing Salesforce. In production: 5,917 outbound messaging rows logged May–August 2026.
Read the storyFassadenklar
A Pipedrive deal becomes a building measured from public 3D city data and a draft quote in sevDesk. No site visit, and no model writes a price.
Read the storyLive in the Movent
Our loyalty, gamification and ticketing platform for live events. We built it with Claude and Claude Code and run its marketing with them too.
Read the story
How it works
How a project runs
From the first call to a system in production.
Call
Thirty minutes about one process: how it runs today, where the time goes, which systems it touches.
Scope
We write down what the system does, what it connects to and what it must never do. If you want this on paper first, that is the AI Systems Audit.
Build
We build it into the tools you already use. A person reviews before anything reaches a customer.
Run
We operate it in production: monitoring, fixes and logs, for as long as it runs.
Free starter
Loopwise Starter
The sanitised version of our own working setup — the same blueprint, rules, and memory system we use to run agents day to day.
Free. After you sign up we'll also invite you to our monthly live workshop — 60 minutes, held in German.
A project blueprint — the folder structure and config we start every build from.
A rules file, CLAUDE.md-style — the constraints we give an agent before it touches code.
A memory setup — how the agent keeps context across sessions instead of starting from zero.
Hooks, skills, and agent definitions — the pieces that turn a chat window into a working setup.
Use cases
Where this shows up
Document processing
Invoices, contracts, intake forms — read, structured, and pushed into the system your team already uses.
Content & AI visibility
Content and product data structured so search engines and AI assistants can find it, read it, and cite it.
Outreach
Messaging and call layers for contacts who opted in, built into your CRM, with every attempt logged against the record it belongs to.
Dental practices
Patient messaging, appointment confirmations and a voice layer for patients who opted in, built around the CRM the practice already runs.
Tax advisory
Email triage, document extraction and client communication, on an environment isolated per firm.
Who builds it
The people who run it
Jan Zajfert
Co-Founder & Business Strategy
Patrick Nemeth
Co-Founder & Technology Lead
Edris Kamal
Business Development
Tell us which process eats your week.
A 30-minute call. We look at the process and tell you straight whether a system is worth building — or point you somewhere else.
Prefer a fixed scope first? Book the AI Systems Audit instead.
Questions we get asked
What does AILoopwise do?
AILoopwise is an AI implementation company. We build and run AI systems for businesses: agents that draft and route work, integrations into the tools you already use, document processing, content and outreach systems. We write the code and operate it in production. We do not sell advice or slide decks.
How is this different from ChatGPT or Copilot?
Those are general tools your team has to operate. We build a system around one of your processes: connected to your CRM, accounting software or inbox, with fixed rules for what the model may and may not do, and a person reviewing before anything reaches a customer.
Which AI models do you build on?
On Claude and other models, chosen per system. In the Fassadenklar build, for example, Gemini and Claude cross-check each other on uncertain measurements, and no model writes a price.
How does a project start?
With a 30-minute call about one process. If a system is worth building, we scope it directly or start with the AI Systems Audit: a review of your processes and data, and a written build plan with an effort estimate.
How long does a build take?
That depends on the process and the systems it touches, so we do not promise a number before we have seen it. One reference point: at Fassadenklar, after we swapped a blocked vendor for sevDesk, the first draft quotes came out of the system the same week.
How is data protection handled?
Every client gets their own isolated system. Where it runs and which model provider processes the data is decided with you for each project, including EU hosting where you need it. We document which part runs where instead of promising blanket compliance.
Which systems can you connect to?
Anything with an API or a webhook. Systems we have connected in production include Salesforce, Pipedrive and sevDesk.
What does it cost?
There is no price list. Every system is scoped to the process it replaces, so the honest answer comes after the first call, not before it.
Do you run the system after launch?
Yes. Building it is half the job. We monitor it, fix what breaks and keep the logs, so a failure is noticed instead of silently piling up.