Industrial digital & IT leadership · Research · Essays
I lead digitalisation and IT at an industrial manufacturing site and sit on its management team. The job is to turn data, software and AI into things operations and finance can count on: a forecast people trust, a plant that sees its own losses, decisions that have an owner.
I came to management from industrial research, with a doctorate from TU Wien and a publication record on maintenance, knowledge and AI in production. In my own time I run a small operating model of AI agents, and write about what it teaches a manager.
Research figures: Google Scholar, November 2026. Operating-model figures: own logs for the 28 days to mid-November 2026. The monthly cost excludes the owner’s time, which is the point of the second essay.
What I work on
Digital programmes seldom fail on technology. They stall on what to standardise, who owns the result and how much change a site can absorb. I run the portfolio accordingly: enterprise systems, production data and planning, each tied to a result that someone outside IT answers for.
Once software acts rather than answers, someone has to say who may decide what. I work on the unglamorous parts: decision rights, approvals, cost tiers, recovery and memory. Governance is a constraint on the design here, never the subject.
A doctorate and 18 publications on maintenance, knowledge systems and AI in production. The research is the foundation of the practice. It is why I ask for evidence before a pilot and for a measure after it.
The case
For about a year, in my spare time, I have run a small operating model: four AI agents with separate jobs, 28 services beneath them and one person accountable for all of it. One of its users is a real small business, whose outbound email waits for a thumb on my phone.
I treat it as a wind tunnel. It shows cleanly where the decisions sit, what machine work costs and how such systems fail, because nothing else gets in the way. It says nothing about works councils, safety regulation or an ageing ERP system, and I try not to pretend otherwise.
What it has taught me so far is unfashionably old. The hard problems were delegation, signing limits, incident reviews and handover notes. Any plant manager would recognise them.
The firm of one is a private project, run in private time on private equipment. It touches no employer data.
Monthly running cost of the whole operating model, in euros
Share of AI traffic by model tier, 28 days to mid-November 2026 (1.56bn tokens, nearly 15,000 calls)
Source: own logs and invoices. About 95% of the tokens were cached re-reads, which is also why they were cheap.
Essays · first essay planned for December 2026
Essays on what a manager must still decide, approve, pay for and remember when work becomes cheap. Each opens with a management problem and closes with a decision that could be taken on Monday.
Four AI agents, one manager and the oldest problem in business
Once agents act rather than answer, a single person is running an organisation, and organisations need management, not better prompts.
Machine labour is nearly free. The bill has moved elsewhere
The marginal cost of machine work has collapsed, so cost control moves from price per unit to capacity, tiering and attention.
I named my agents. The research says that was a mistake
An agent is equipment with an owner, not a colleague with a role; the framing changes how carefully humans check.
As of November 2026
Enterprise systems, production data and applied AI, with the plant running on Monday as the fixed constraint.
A series of essays from December 2026, roughly one a month, with the odd month off for good behaviour.
A design-research paper in preparation. I am looking for an academic co-author with design-science experience. Details
TU Wien, Institute of Management Science.
Preferably in front of operations and general managers. Topics and formats
Contact
A mandate where digital work has to show up in operating results. A stage or a page for one of the essays. A paper that needs a practitioner with data.