AI & Automation
AI and automation where
it actually pays off.
Automate repetitive tasks and integrate AI meaningfully into your processes, without hype, with real value.
Discuss your projectFree initial conversation · Directly with Daniel Gläser
- 01Input
- 02AI draft
- 03Human review
- 04Approved action
Assess the value
AI is not an end in itself
A recurring, clearly described process is a good starting point. We compare effort, frequency and the cost of errors. Fixed rules often only need conventional automation; AI can help with unstructured text or knowledge search.
A pilot shows whether quality and time savings hold up in daily work. Model, hosting, maintenance and human review costs belong in the calculation from the start.
Use cases
What I offer
Workflow Automation
I connect your systems into workflows that run in the background. Syncing data between apps, processing emails, generating reports, without anyone kicking it off by hand. Which automation platform sits underneath depends on your setup.
AI Chatbots & Assistants
Intelligent bots that answer customer inquiries or search internal knowledge bases. Based on modern language models, grounded in your knowledge base (RAG).
AI Email Agents
Classify messages, prepare relevant draft replies and route exceptions to people. Automatic sending is configured only for explicitly approved workflows.
Data Analysis with AI
Evaluate large amounts of data, recognize patterns and create reports. From customer feedback analysis to forecasting.
Agents & integration
Find knowledge. Use tools deliberately.
RAG
Assistants search your approved documents and knowledge base. Source references help traceability; answers still need checking.
Function Calling / MCP
Defined interfaces connect language models to your systems. Permissions limit what an agent may read, change or trigger.
Documents, email & phone
Extract invoices and forms, triage messages or receive and route calls. Customer onboarding, data synchronisation and reports can also be linked in controlled workflows.
Data & control
Hosting is one part of the design.
EU-hosted or local models are selected for data sensitivity, performance and operating cost. Data flows, providers, retention, access and required agreements are reviewed for the specific use. Location alone does not guarantee GDPR compliance.
We review transparency and labelling duties for the application and define when people approve results. Consequential actions need clear boundaries, logs and a manual handling path. AI can produce incorrect answers; testing and ongoing quality review are part of operations.
Working together
A clear path to everyday use.
Understand the process
Record the starting point, data, effort and a measurable goal.
Build a small pilot
Test quality, limits and costs on representative cases.
Integrate with safeguards
Define permissions, approvals and error paths; connect existing systems.
Monitor quality
Review outputs, assess feedback and improve specific weaknesses.
Before we begin
Good questions. Clear answers.
Do I need large datasets?
Not necessarily. An assistant can start with a small, maintained knowledge base. Relevant content, permissions and representative test cases matter more.
Does the AI decide independently?
Only within explicitly defined tasks and permissions. Consequential actions have approval and escalation paths. Uncertain results go to a person.
Does EU hosting automatically ensure data protection compliance?
No. Legal basis, data flows, providers, access and contractual duties need to fit the specific use and are reviewed during the project.
What does an AI project cost?
Scope, integrations, model usage and review effort determine cost. We begin with a bounded pilot and agree its effort before implementation.