DIGITALISATION & AI
AI rarely fails on the technology
Advertised as: Digitalisation Project Manager · ERP Project Manager · AI Governance Consultant
It fails on data no one has sorted, on processes that were never described, and on a missing framework for use. Introducing tools before those three things are in place creates effort without result – and, in case of doubt, a data protection problem.
I work in the reverse order.
The sequence
First the use cases.
Not "where could we use AI", but which process costs time today, how often it runs, what an error there means. Some cases are worth it, many are not. Saying both is part of the job.
Then the data.
Cleansing, structuring, modelling. A language model on an unsorted drive delivers confident wrong answers – that is worse than no answer. I have been doing data modelling and migration in ERP projects for twenty years; the task is not new, only the occasion.
Then the repositories.
Building knowledge bases and document structures so that they are machine-usable and factually correct. That includes the uncomfortable question of which documents still apply at all.
Then the framework.
Who may do what with which data, where the approval paths sit, how a result can be verified. GDPR implications and the forthcoming obligations of the EU AI Act taken into account before the first process goes live.
Then the tools.
Only at this point. Automation with n8n, AI-supported workflows – set up for a process that someone has understood beforehand.
And finally the people.
Defined use cases, clear boundaries, verifiable results. Tools without enablement stop being used after three weeks.
Where this comes from
The digitalisation part of this work is not a new discipline for me. I have been doing CRM and ERP implementations and migrations for mid-sized companies for twenty years – ERPframe and DOCUframe, ERPnext, Dolibarr, Odoo, Equinox. Data modelling, data cleansing, migration, interfaces via EDI and to web shops, document management.
Results from that practice: sales processes roughly 40 % faster, technical service roughly 50 % more efficient, and in the reorganisation of one company more than 30 % savings in personnel costs through process optimisation. That was long before AI and with the same underlying questions: which process actually runs, which data is reliable, who works with it.
On the AI side I am a practitioner, not a researcher. I have implemented my own automations and process workflows with n8n and AI-supported tools – on a manageable scale, but built myself and in operation.
What I am not
I do not develop models, train nothing and build no machine learning pipelines. If your project needs data science in the narrow sense, I am the wrong person.
My contribution lies before and alongside that: evaluating use cases, preparing data and processes, setting the regulatory framework, steering the implementation and bringing the organisation along. On the technical parts I work with specialists and coordinate them.
AI is my tool. The judgement is mine.
Typical projects
AI roadmap
collect, evaluate and prioritise use cases; estimate effort and benefit realistically; establish the order.
Prepare the data basis
review, cleanse, structure and model existing data before it feeds into AI-supported processes.
Knowledge management
build document repositories and knowledge bases so that they are factually reliable and machine-usable.
Process automation
automate recurring workflows with n8n: document processing, data handovers between systems, notifications, reporting.
Usage framework and governance
policy for the use of AI, roles and approval paths, data protection assessment, preparation for the obligations of the EU AI Act.
CRM, ERP and system replacement
selection, implementation, migration, interfaces. Classic digitalisation, often the prerequisite for everything else.
Framework
AVAILABLE
- SCOPE
- According to project scope; roadmap development also available as a defined short engagement
- DEPLOYMENT
- DACH region, predominantly remote
Where do you stand?
If you are unsure whether AI is worth it in your case at all, that is a good starting point for a conversation. I will tell you plainly if the answer is no.