Start with a concrete workflow.
Reduce recurring work, connect existing systems and check results against clear criteria. I analyse the process and develop a suitable combination of software, interfaces and AI. Feasibility depends on data quality, business rules and the consequences of errors.
Prepare incoming inquiries
Classify messages, extract required information and prepare replies for human review. Incomplete or ambiguous inquiries go to a responsible person.
Process documents
Extract information from recurring document types, check mandatory fields and prepare structured data for downstream applications. Exceptions receive a separate review path. For fixed rules and structured data, a reliable script may be sufficient without an AI model.
Automation assessment
Map the workflow, systems, data and dependencies. Identify opportunities, constraints and a prioritised first process. Agree scope, deliverables and fees before the assessment.
A focused pilot
Build a bounded solution and compare quality, processing time, rework and running costs with the current approach. Test representative cases, approvals and fallbacks before deciding on deployment.
Implementation and ongoing support
Integrate the tested solution, document operation and responsibilities, train users and agree maintenance. Each phase can be commissioned separately; effort and timing depend on the process and interfaces.
Plan quality and operation from the start.
Measure benefit, agree data access and suitable local or external models, retain human control and document maintenance. Confidential information is not sent to external AI services without an agreed basis. Critical results remain subject to professional review. Read about the approach.
Custom AI assistants. Controlled data flows.
Make documents searchable and support digital work through a tailored assistant connecting models, a web interface and existing applications. Test functions and output quality on actual tasks. Self-hosted models can run in your own or dedicated infrastructure for confidentiality requirements; permissions, permitted connections and support are agreed according to risk.
Implemented example · Protection requirements and operating models
Software, data and operations.
The following services support automation projects and can also be commissioned independently.
Custom software and automation
Develop tools, services and integrations for specific business requirements. Analyse existing applications, plan architecture and implement maintainable solutions with tests and documentation.
Data integration and migration
Connect applications, validate imports and exports, map fields and migrate existing data using repeatable procedures. Define the authoritative system and make conflicts and failures visible.
Linux, infrastructure and operations
Set up and maintain agreed systems and services, containers, monitoring, backups and migration procedures. Document dependencies and restoration paths.
Websites and online shops
Build, modernise and maintain websites and specialised commerce processes with WordPress, WooCommerce, PHP, HTML and CSS. Include data migration, roles, interfaces, performance and technical accessibility.
Technical advice and design
Assess systems, risks and options before changes. Produce understandable architecture, security, backup and operating concepts with priorities, expected effort and explicit assumptions.
Technical quality and evidence
Review agreed technical criteria for websites, shops and servers. Document findings, risks, tests and corrective actions. Formal certification and legal assessments are outside this technical review.
Support with clear agreements.
Agree scope, availability, response times and escalation before work begins. Continuous availability or a 24/7 incident service is not part of the general offer. Expand automation after the first workflow has proved itself. Technologies and expertise.
Questions before the first project
Where should the first project start? Choose a recurring, bounded process with representative data. Establish processing time, common errors and approvals before the pilot.
Must existing software be replaced? Start with current systems and check interfaces, exports and complementary tools. Replacement is considered only where the technical options and agreed objective warrant it.
When is AI useful? Fixed rules and structured data often suit ordinary software. AI can help classify variable texts or extract document fields. Uncertain results need human review.
Linux, hosting and backups
Configure and maintain Linux, domains, DNS and SSL/TLS. Plan backup and restoration, test recovery and document migrations and server transitions. The objective is understandable infrastructure and reproducible recovery.