AI process automation and optimisation
Automate recurring workflows with AI that fits your business
Invoices, orders, enquiries, reports: much of it is still handled manually. I analyse your workflows, find the steps with the biggest leverage and automate them with a mix of classic software and AI models.
Most businesses have workflows that eat time every day: sorting and answering e-mails, typing in receipts, assembling quotes from templates, copying data from one system to another. Classic automation often fails here because the input is unstructured: free text, PDFs, photos, phone calls.
That is exactly where AI models help. They understand text and documents and can prepare decisions. Combined with robust software you get a workflow that runs reliably while keeping your people in the loop: the AI suggests, a human decides where it matters.
What you get
-
Hours per week back
Routine work runs in the background. Your team looks after customers instead of typing.
-
Fewer errors
Automatic checks catch transposed digits, missing details and duplicates before they cost money.
-
Humans stay in control
Critical steps are only prepared, never executed automatically. Approvals stay with you.
-
Fits your existing systems
Connects to the software you already use (ERP, accounting, CRM, e-mail, DATEV, Lexware, Excel) instead of adding yet another isolated tool.
Understandable real-world examples
Capturing incoming invoices automatically
- Starting point
- A trade business with 25 employees receives around 300 supplier invoices a month by e-mail and post. An office clerk types invoice number, amounts and line items into the accounting system.
- Solution
- A mailbox robot collects attachments, an AI reads the invoice data (even with varying layouts), matches it against purchase orders and creates a booking proposal. Unclear cases go to a review list.
- Result
- Data entry reduced from about two days a month to a few hours. Review stays with a human, the typing is gone.
Pre-sorting and answering customer enquiries
- Starting point
- An online retailer receives 80 e-mails a day: delivery status, returns, product questions. Everything lands in one shared inbox.
- Solution
- An AI classifies each enquiry, extracts order number and request and drafts a reply using data from the shop system. Standard cases are sent after approval, special cases are routed to the right colleague.
- Result
- Response time down from two days to a few hours. The team only answers the cases that really need attention.
Weekly report at the push of a button
- Starting point
- A manufacturing company compiles a report every Friday from Excel lists, machine data and notes. Duration: half a day.
- Solution
- A script collects the data from the sources, an AI summarises anomalies in plain sentences, the result goes to management as a PDF.
- Result
- The report is ready at 7 a.m. on Friday without manual gathering. Anomalies are visible at a glance.
How we work together
-
1
Free initial consultation
30 minutes by phone, video or on site in and around Eisenach. You describe what costs time or causes friction today. I tell you honestly whether and how AI can help.
-
2
Analysis and concept
I look at the workflow, the data and the systems involved. You receive a short concept with effort, benefit and risks in plain language.
-
3
Implementation in small steps
First a working prototype, then the build-out. You see results early and keep control over budget and direction.
-
4
Rollout and support
Training for your team, documentation, monitoring. On request I take care of maintenance and further development.
Frequently asked questions
- Do our data have to go to the cloud?
- Not necessarily. Depending on the requirements I use European providers, data processing agreements or locally hosted models on your server. We settle this in the concept phase before a single line of code is written.
- How quickly will we see results?
- A first working prototype for a clearly defined workflow is usually possible within two to four weeks. After that you decide whether and how to expand.
- What happens when the AI gets it wrong?
- Automation is built so that uncertain cases are detected and handed to a person. Critical steps such as payments or shipping are only prepared, never executed without approval.
- Do we need new software?
- Usually not. I build bridges between the systems you use today. Only if a system blocks automation do we talk about alternatives.
Related services
Data extraction with AI
Invoices, delivery notes, contracts, forms, product catalogues: valuable information sits in documents nobody wants to type up. AI-assisted extraction pulls it out reliably and hands it to your systems in a structured form.
Learn moreSoftware development with AI techniques
Many years of web development experience with Ruby on Rails meet modern AI tooling. The result: tailored applications that take weeks instead of months and include AI features where they deliver real value.
Learn moreMCP development: connecting AI assistants to your systems
The Model Context Protocol (MCP) is the open standard that lets AI assistants like Claude or ChatGPT access company systems: calendar, inventory, CRM, databases, line-of-business applications. I develop MCP servers that expose exactly the functions you want to expose.
Learn moreLet's talk about your use case
In a free 30-minute call we find out whether it is worth it and what a first step could look like.
On site for businesses in Eisenach and within 100 km, remote beyond that.