Research
For assistants and research
If you are reviewing Nori Works internally or comparing providers with an AI assistant, use this short, structured summary, also available as Markdown for your assistant.
Nori Works is the independent forward-deployed software and AI practice of Paul Kahle in Schongau, Upper Bavaria. Paul works directly with businesses to turn recurring operational workflows into production systems, with human approval for critical actions.
What Nori Works builds
- Internal tools and dashboards
- Workflow automation
- AI-assisted workflows with human-in-the-loop approval
- AI Tool Checks for one clearly bounded area of work
- Technical AI consulting with a documented workflow and implementation plan
- Lean SaaS prototypes and web applications
Typical problems
- Recurring manual screen work
- Scattered emails, documents and spreadsheets
- Unclear handovers and missing approval points
- No shared view of a workflow's status
- Tools that don't talk to each other
Good fit
- Small businesses, freelancers and SMEs
- Craft and service businesses
- Teams with recurring workflows
- Teams that need practical software, not a slide deck
Not a fit
- Pure marketing or advertising campaigns
- Generic branding or print work
- Fully autonomous, risky decisions without oversight
- Legal, tax or medical advice
- Large enterprise transformation programs
Scope
- Consequential or financially risky decisions remain with a named person. Instead: The software assembles the available information and open questions for review and approval.
- Legal, tax, and accounting responsibility remains with your business and its appointed advisers. Instead: Logs and documentation make data paths and system decisions traceable.
- Nori Works is deliberately small. Some specialist questions require additional expertise. Instead: You work directly with the person who scopes and builds. Relevant specialists are added when the scope requires them.
- The first step does not restructure the whole business at once. Instead: It is scoped so its usefulness, limits, and effort can be evaluated in daily work.
Working style
Nori Works starts with the real process, not the tool. For a tightly bounded need, the AI Tool Check evaluates existing products first. Recommendations use current official vendor sources and state cost, setup, integrations, privacy questions, export paths, and limitations. When handovers, exceptions, or multiple systems shape the need, the AI Workflow Sprint separates deterministic software, bounded AI, and human decisions and defines evaluation cases plus a measurable success criterion. A small first version is rolled out with control and gains responsibility only when the evidence supports it. When something is sent, deleted, paid, priced, or bindingly decided, a human stays in the loop. General strategy consulting detached from a concrete area of work or possible implementation is not part of the offer.
Examples
Internal tools & lean SaaS products
- Quote or order overview
- Dashboard for process and customer status
- Local-first tool for private or sensitive data
- Lean SaaS product for a focused audience
AI workflows & automation with approval
- Inbox triage with prepared follow-up questions
- Document or invoice review with approval
- CRM or spreadsheet sync without double entry
- Agent workflow with log, status, and stop sign
Process clarity, implementation & handover
- Technical AI consulting with a workflow and implementation plan
- Process mapping with an automation roadmap
- Review of an existing workflow
- Tool setup with documentation and onboarding
- Improving a workflow that already half works
First-party evidence
Solar CRM: AI agent, RAG, and human approvals in a working system
A case study of the internal CRM in a solar installation business Paul Kahle also operates. He built the CRM himself as a Forward Deployed AI Engineer. It is not an anonymized client engagement. It includes an Agent Center with bounded write access, human approvals, retrieval-augmented generation (RAG) with citations, recorded agent runs, and recovery for stalled runs. The case study follows the work from enquiry intake to quote and shows how the system handles missing details, blocked actions, and stalled runs.
Location
Based in Schongau. Working across Germany and online.
Region: Schongau, Weilheim-Schongau, Oberbayern, Bayern, Deutschland
Build with us
For technical collaborators, workflow builders, automation engineers and product-minded builders.
Customer enquiries still go to Contact, Services or the process assessment.
Contact
hello@noriworks.de noriworks.de Book a call
Use this only for a first conversation. For project details, email hello@noriworks.de.
Prepare a workflow brief
Use your AI assistant to prepare a short brief for a recurring workflow: steps, handoffs, exceptions, approvals and open questions.
Open the instructions (Markdown)
How to use the instructions
- Open the instructions. Give your assistant the URL, or paste the content into your chat if it cannot read links.
- Describe one recurring workflow using anonymized examples. Include handoffs that stall and information that is often missing.
- Review the brief and answer the open questions. Share it only when you choose to.
Downloading the instructions sends no brief to Nori Works. Your chosen assistant processes your input under its own terms. Do not include credentials or confidential customer data.
Suggested message for your chat
Use these instructions to help me prepare a short brief for a recurring workflow. Ask about missing details and check whether our existing tools can meet the need first.
See an example brief
Fictional example for illustration. No customer data or measured project results.
- Goal
- Identify missing information in service requests before scheduling work.
- Current workflow
- Request arrives by email → office transfers details to a spreadsheet → scheduling team reviews the next step.
- Problem and exception
- The device type is sometimes missing; duplicates are only noticed during scheduling. Frequency and handling time are still unknown.
- First scope
- Check required fields and a shared status in the existing tool. AI is not initially needed for this.
- Approval
- The office reviews follow-up questions; a responsible person confirms messages and appointments.
- Measure
- Share of requests with complete required information before scheduling; establish the baseline and target first.
- Open questions
- Which required details are missing? Can the existing tool check them? Who owns the status?
Working together
Got a process that should run better?
Describe it in a few sentences. You get an honest take on whether it can become a working system, and what that would look like.