Case studies
More than 25 solutions delivered
They include complete applications, individual modules, interfaces and automated checks, running for clients and in our own products. We do not name our clients. Nine examples follow below, each with its task and result.

You will be speaking with Ing. Thomas Grafenau, BSc MSc. He leads the intro call and builds your software.
12 years of software development for mid-sized companies
More than 20,000 automated tests for business software that had grown over years
- Client
- Mid-sized industrial company with several sites
- Starting point
- The business software had grown over many years. There were practically no automated tests. Every change could only be checked by hand.
- Task
- Build a test base after the fact without halting ongoing development.
- Approach
- New tests run red first; only then is the code written. We secure existing code like this: we deliberately set each check to a wrong value once and watch it fail. Tests and build run before every delivery. On top of that come our own check scripts, for example for permitted dependencies between modules. The number of open lint findings may only go down.
- Result
- In around ten weeks, the number of automated test cases rose from almost zero to more than 20,000. Since then, new features have entered the system together with their tests.
AI agents with an orchestrator: a new sales module in five days
- Client
- Mid-sized industrial company with several sites
- Starting point
- Sales needed a module to record new business contacts and distribute them within the team. A clickable prototype with test data already existed.
- Task
- Turn the prototype into production code with a real data connection, backed by tests.
- Approach
- The basis is a written specification. An orchestrator agent breaks it down into work packages and distributes them to specialised AI agents. For each package, the tests are written before the code. Reviewer agents that were not involved in the build check the results. We approve the plan and sign off the result.
- Result
- Five days passed from the finished specification to delivery to the production system. The effort was around 60 working hours, including preparing the acceptance test. The module comes with around 3,800 automated tests. An AI agent was involved as co-author in 98% of the changes to this software since September 2026.
Field service scheduling with a tablet view for technicians
- Client
- Mid-sized industrial company with several sites
- Starting point
- Field jobs were planned separately from the rest of the company data. Planning was to move into the existing business software.
- Task
- A dispatch tool with calendar and map. In addition, a tablet view for technicians that automatically produces the job report on completion.
- Approach
- We described the requirements in functional specifications and built in short stages. Dispatch sees travel times between jobs. On site, the technician documents the work with photos and has the customer sign on the tablet. This produces the report as a PDF. Customers are notified by text message.
- Result
- After around two months, field service scheduling went live on the production system. Since then, around 60 further tasks have been completed. More than 4,000 automated tests cover the module.
Route and maintenance planning for service companies
- Client
- Our own product for service companies
- Starting point
- Service companies carry out regular maintenance at their customers' sites. Every maintenance visit must be documented in a traceable way.
- Task
- A platform for weekly and route planning with route optimisation. In addition, an app for technicians that also works offline, and maintenance reports with signature as PDF.
- Approach
- The first version went into production after around one third of the effort spent so far. We then extended the platform in short steps based on user feedback. It supports multiple tenants and has online payment so that several companies can use it. More than 200 automated end-to-end tests check the workflows in the browser.
- Result
- The platform is running in production. Around 220 tasks have been completed so far.
CRM with ERP integration for sales
- Client
- Mid-sized industrial company with several sites
- Starting point
- Customer data and transactions were held in a legacy system that was to be replaced. Documents and line items were held in the ERP.
- Task
- A web-based CRM for the sales team and its managers, connected to the ERP and the website.
- Approach
- An automatic synchronisation fetches the documents from the ERP. Quotes are created directly in the CRM from this data. For each order, the system calculates the contribution margin. Enquiries from the website automatically create a customer and a task. Managers maintain their own sales targets.
- Result
- The CRM is in use and is gradually replacing the legacy system. Nobody has to retype documents from the ERP any more.
Learning platform for onboarding and mandatory courses
- Client
- Mid-sized industrial company with several sites
- Starting point
- New employees were to be onboarded according to a fixed plan. Mandatory courses needed deadlines and proof of completion.
- Task
- A learning platform in which managers assign courses and see their team's progress.
- Approach
- Written functional specifications form the basis. Alongside the course catalogue, there are onboarding plans for new employees. Mandatory courses have a deadline, and managers assign them. Appointments go out as calendar invitations, reminders arrive by email. Reports show learning progress.
- Result
- The platform has been activated on the production system. Build time so far is around 140 hours. More than 4,300 automated tests cover it.
Digital production control
- Client
- Mid-sized manufacturing company
- Starting point
- Manufacturing was to be controlled digitally from end to end. Data from the design software was to flow directly into production.
- Task
- Production software from order intake down to the individual workstation. Production management gets its own view.
- Approach
- The design data is imported. Capacity is visible per calendar week. Permissions are assigned per role. Label printing and digital quality control are built in. Sign-in uses the existing company account. We built it in stages.
- Result
- Manufacturing runs live on the system. Around 230 features, changes and bug fixes have been implemented.
AI guide for the workforce of a large enterprise
- Client
- Large enterprise
- Starting point
- The company provides the AI coding tool Claude Code to many employees. It was meant to be used widely and productively.
- Task
- A practical guide in German and English. It is worded neutrally and contains no internal names.
- Approach
- Chapters with short sections and videos, maintained strictly in parallel in two languages. Access is protected.
- Result
- The guide is online and is being extended continuously.
Tom's Clock: time tracking as our own product
- Client
- Our own product
- Starting point
- We record our own project hours per client and project. For this, we built our own time tracking.
- Task
- Multi-tenant time tracking with a time clock and project times, as a software subscription with a free and a Pro tier.
- Approach
- We use Tom's Clock in our own business and have extended it alongside. Leave requests and online payment were added, as well as a Pomodoro timer. This extension took around 72 hours.
- Result
- In our own business, we recorded 2,011 hours in 1,246 entries in just under six months. Tom's Clock is publicly available.
From our own business
We built this for our own company, not for a client.
AI agents in our own office: matching receipts, working through the inbox
- Client
- Our own business, not a client project
- Starting point
- Every month, the card transactions in our accounting have to match their receipts. Many kinds of email arrive in our inbox, and each kind needs a different response.
- Task
- The goal: two AI agents take over this recurring office work. A person still approves anything that gets posted, and anything an agent does not recognise goes to a person.
- Approach
- The accounting agent matches every card transaction to its receipt and produces the month's Excel list. If a receipt is missing, it creates a draft in the accounting software; it cannot finalise it. The inbox agent sorts incoming emails by rules defined for each kind of email. If it does not know a kind, it leaves the email where it is and asks.
- Result
- Both agents are in use in our business. In accounting, drafts are created instead of posted entries; a person checks and approves them.
Intro call
Let's talk about your project.
30 minutes, free and without obligation.
Pick a time
30 minutes, free, no obligation
Prefer to talk directly? Call or write:+43 664 4128491office@grafenau.at