Where AI actually shows up in your daily work
AI doesn't work as a tool, but within a task. That's why we sort by department rather than by function: what costs time there today, which example concretely solves it – and what to watch out for. Choose the department closest to yours.
The classic office
Invoices, orders, master data, filing: work that never runs out, and whose effort consists almost entirely of searching, transcribing and following up.
- Typing invoice data out of PDFs and matching it against orders
- Chasing approvals by email while early-payment discounts lapse
- Searching sprawling file systems for the current version
- Sorting recurring requests in a shared inbox
Incoming invoices
Read the document, match it against the order, flag anomalies such as changed bank details, and submit it for approval – posting is still done by accounting.
Sorting the inbox
Classify incoming messages by request type, route them to the right team, and attach a draft reply for standard cases.
Answers from the files
“What terms did we agree with this supplier?” – answered from your own documents, with a source reference.
Sales, proposals and tenders
Proposals are rarely written from scratch – they're assembled from earlier proposals. The effort lies in finding things again, not in making the case.
- Hunting down the right text blocks from old proposals
- Combing tender documents for required evidence
- Reconstructing where a case stands after a follow-up question
- Proposals that never get written at all for lack of time
Proposal draft
Generate a rough draft from approved building blocks and comparable earlier proposals – prices come from the costing, not from the text.
Breaking down a tender
Create a structured list of requirements and deadlines from the bundle of documents, which also serves as a checklist before submission.
Consistency check
Before sending, check whether another customer's name is still in the document or dates contradict each other.
HR department
Text-heavy, and at the same time the legally most sensitive area. The dividing line runs wherever a system influences a decision about a person.
- The same questions about vacation, travel expenses and parental leave, several times a day
- Assembling job ads from old templates
- Building onboarding plans from scratch for every role
- Reference-letter drafts in the usual formulaic language
Employee self-service
An assistant answers questions about internal rules from the current documents – with a source, and a note that HR decides individual cases.
Drafting a job ad
Generate an ad from the task description and requirements profile, and check it for wording that could be off-putting or relevant under German anti-discrimination law.
Preparing onboarding
Derive a structured onboarding plan per role from existing documents.
Candidate selection and performance evaluation count as high-risk applications under Annex III of the EU AI Act – full requirements apply from 2 December 2027. We deliberately advise caution here.
Municipalities, authorities and public enterprises
A heavy volume of text, staff shortages, and special requirements for proper record-keeping, traceability and data sovereignty.
- Knowledge that walks out the door with every retirement
- Meeting papers and minutes from extensive attachments
- Recurring citizen inquiries by phone
- Translating official notices into plain language
Research in your own records
Make past cases, bylaws and policies searchable – with a source reference and within the relevant access permissions.
Meeting papers
Create a draft in the standard in-house structure from the current status and history, and summarize long attachments for committees.
Plain language
Translate official notices and information texts into plain-language or multilingual versions – with expert review before publication.
Processing within Germany, involving the staff council, and a clear separation between preparation and decision are not optional here — they're a prerequisite.
Developers, engineers and design
The biggest lever isn't code generation, but documentation and the experience-based knowledge that's written down nowhere.
- Understanding someone else's code or old designs
- Documentation that's chronically behind
- Searching standards and internal guidelines
- Fault patterns a colleague already solved years ago
Unlocking experience-based knowledge
Make service reports, tickets and project files searchable: “Have we seen this fault pattern before on this type of system?”
Documentation from what already exists
Generate a draft from change logs and ticket histories that a specialist corrects – instead of writing from scratch.
Standards and guidelines
Query internal policies and regulations with source references, compare datasheets and parts lists for discrepancies.
Manufacturing, service and maintenance
In shift operations, what counts is how fast the right information is available – and whether anyone can find it at night too.
- Work instructions and inspection plans scattered across different repositories
- Fault reports with no link to earlier cases
- Training new colleagues ties up experienced staff
- Checking supplier documents for completeness
Assistant for the shop floor
Answer questions on work instructions, inspection plans and safety rules – with a reference to the current document.
Fault history
Match a current fault pattern against earlier cases and show the fix that worked back then.
Preparing reports
Generate a draft inspection and acceptance report from measurement values and log notes.
Marketing
The first area AI arrived in, and the first where disillusionment set in. The difference lies in whether the system actually knows your substance.
- Text that sounds like everyone else's
- Expertise sits with people who don't enjoy writing
- One piece of content has to be adapted for five channels
- Topic planning by gut feeling instead of by demand
Putting expertise to use
A recorded conversation with a technician becomes the basis for an article with genuinely distinctive content.
Topics from real questions
Derive which questions keep recurring from customer inquiries and support tickets – and build the editorial calendar from that.
Channel adaptation
Turn a vetted expert article into a short version, a newsletter paragraph and a video script.
The transparency obligations under Article 50 of the EU AI Act have applied since 2 August 2026 – chatbots need a notice, AI-generated content needs a label.
Management and IT responsibility
You're not deciding on a single use case, but on the framework, sequence and accountability.
- Pilot projects that stall after the workshop
- Shadow AI that nobody knows who's using
- Unclear legal situation and shifting deadlines
- Cost questions with no solid benefit calculation
Stocktaking
Record which AI tools are actually in use, and derive an approved alternative from that which beats the private account.
Setting the framework
Usage policy, Article 4 competence evidence, processing location and data processing agreement – in a few days, not months.
First case with a measurement point
Pick one use case, run it cleanly for twelve weeks, and make a solid decision at the end: scale it up, refine it, or stop.
Which case fits you?
In the initial conversation, we look at a concrete process from your organization – including the edge cases no process description ever mentions. You'll get an honest assessment, even if it's that AI wouldn't pay off here.
Your secure AI platform for the Mittelstand. Secure. Intelligent. Integrated. Custom database integration, personally supported.
novendix GmbH · Industriestraße 6 · 91126 Schwabach
Locations: Schwabach · Weißenburg · Nuremberg
A company of the L&S Lange & Schermer Group
