Production AI systems, with the evidence they work.
re-entry builds the AI systems that turn inquiries, documents and emails into work done in your ERP and CRM: quotes drafted, invoices posted, tickets routed. Every build ships with the eval suite that proves it, a person on every decision the system is unsure about, and an audit trail of every change. Fixed price, first demo inside two weeks, EU servers or your own, and the code is yours.
The factory runs a limited number of build lanes. Join the waitlist and we email you when the next one opens.
Quotes
Müller GmbH
Inquiry: 40 pallet racks, delivery in May
Pallet rack PR-90 × 40
€18,400
Assembly, 2 days
€2,300
Total
€20,700
Concept screens from the catalogue: a quoting tool, a document inbox, an assistant on your files. Yours is built to your workflow.
Why AI pilots stall before production
Most AI projects stall between the demo and daily use, and the model is rarely the reason. The inputs are PDFs, emails and spreadsheets that no two suppliers format alike: Gartner expects organisations to abandon 60 percent of AI projects that lack AI-ready data, and in BARC's 2026 study 70 percent of data leaders said less than half of their unstructured data is findable and usable for AI. Deloitte's Tech Trends 2026 counts 11 percent of organisations running agentic AI in production.
Coding agents make implementation cheap. What stays expensive is proving the system still works after the next change, and that is where our engineers spend their time: they write the data model, the acceptance tests and the eval cases from a sample of your real documents before an agent writes a line. Supervised agents produce the implementation, and every merge has to pass the tests and an engineering review. A German agency bills €90 to €140 an hour over a quarter or two; the factory ships in weeks, at a price fixed before work starts.
The last wave burned people. S&P Global found that 42 percent of companies scrapped most of their AI initiatives in 2025, and in VentureBeat's 2026 agent reliability tracker half of the organisations surveyed had shipped an agent that passed internal evals and still failed in front of customers.
So the evidence is part of the delivery: typed code, tests, the eval suite run against your documents, a person on every decision the system is unsure about, and a named senior engineer signing every release.
What we build
Documents into your ERP
Invoices, orders, delivery notes and contracts read, checked against your master data and posted to your ERP. Anything the system is unsure about goes to a person with the source document attached.
Quoting and tenders
Apps that read the inquiry, match positions against your catalogue and past offers, and draft the quote for a person to send. Fraunhofer IAIS documents quote preparation dropping from two days to thirty minutes with this class of tool.
Customer service and triage
Email and ticket triage in German and English: sorted, drafted and escalated by rules you set, with a person on every send that matters.
Assistants on your documents
Answers across your wiki, tickets, drives and mail archive, with sources cited and the same access rights your people have. Runs on EU servers or your own, and your data trains nobody's model.
AI in your existing software
Customers ask for AI in your product: we build the feature in your stack instead of rewriting your product. If an AI-built prototype collapsed before production, we rebuild it so it survives daily use.
Customer-facing MVPs
A product idea with a deadline. Auth, payments, admin and deployment, live in weeks and production-grade from the first release.
How a build runs
Week 0
Scope call
Thirty minutes on the problem, then a written offer: scope, fixed price, demo dates. If the factory is the wrong tool for the job, we say so on this call and point you somewhere better.
Week 1
Plan and skeleton
Data model, integrations and a walking skeleton deployed behind a URL you can open. Our engineers write the plan, the acceptance tests and the first eval cases from a sample of your real documents; agents do not write their own exam.
Weeks 2 to 3
Our engineers orchestrate and review
Coding agents, Claude Code among them, work through the plan around the clock. Every change passes the test suite and an engineering review before it lands. You get a demo each Friday and a changelog in plain language.
Week 4
Hardening and handover
Security and failure testing, documentation, deployment to your infrastructure or to EU hosting we manage. The eval suite, test reports and an audit trail of every merge ship with the code.
A week-two build board. Agents produce, tests gate every merge, and our engineers sign the release.
The founder behind this
re-entry is founded by Kyle Nelson, a senior engineer based in Germany whose track record runs from building Debtist's collections platform to a €100 million exit. Before the factory, the team shipped an AI code review platform that seven engineering teams ran against their pull requests in production (retired in 2026), and Fazit, a macOS app that turns calls into finished notes without audio leaving the machine.
The factory is that same pipeline pointed at your problem: the agent orchestration, review gates and test discipline built for our own products now build yours. You talk to the founder, and senior engineers review every merge from the first call to the handover.
See the work: getfazit.com & github.com/Re-entry-ai
Built by this factory
Pricing
Every price is fixed and stated before work starts. All prices plus VAT.
Pilot build
€14,900
three weeks, fixed
One workflow, built end to end. The right start when you want proof before a bigger commitment.
- A working app, deployed and in daily use
- Eval suite built from your own documents
- Repo in your GitHub organisation from day one
- Demo every Friday
- GDPR data processing agreement included
- 30 days of fixes after handover
Production build
Most builds€49,000
six weeks, fixed
The full application: accounts, roles, integrations with your ERP or CRM, and a proper handover.
- Everything in the pilot build
- Auth, roles and admin surface
- Integrations with your existing systems
- EU hosting we manage, or your own infrastructure
- Tests, CI, documentation and an audit trail of every merge
- EU AI Act classification memo for the app
- 90 days of fixes after handover
Factory lane
€7,900
per month, stop any month
A standing lane in the factory: continuous building on a roadmap we shape together every month.
- One active build stream at all times
- Roadmap review monthly, demo weekly
- Maintenance of everything already shipped
- Priority response on production issues
- Pause or stop at the end of any month
If a build stops making sense partway through, you pay for the weeks delivered and keep everything built so far, including the repo.
What to read before a scope call
German articles that answer the questions that stop a budget: price, contract form, evidence, the contract itself, and what the law asks of you.
Frequently asked questions
Start a build
Email two or three sentences about the problem to hello@re-entry.ai. You get a reply within one working day, a thirty-minute scope call if it fits, and a written fixed-price offer inside a week. If the factory is the wrong tool for it, we say so on that call. You can also book the scope call directly and skip the email.