Research
Everything below has a permanent identifier, so it can be checked without going through me.
Peer-reviewed publications
Problems of implementing artificial intelligence in enterprises. Reasons for low return on investment and management solutions
Argues that the barriers to realising a return on enterprise AI are predominantly organisational rather than technical, and sets out a structure for identifying them before a programme starts rather than after it fails.
Research software
ai-payback
An instrument that turns the framework above into something an organisation can run against itself. It scores 78 diagnostic questions across 26 barriers, builds a total cost of ownership covering the categories most often left out of AI budgets, and computes simple payback — or refuses to, and says why.
Its design commitments are enforced by tests rather than asserted in prose: an unanswered
question is never scored as zero, an unpriced cost is never treated as zero, no number is
invented, and every coefficient carries the tier of its source and the date that source
was last checked. Claims that could not be traced to a reachable primary document are
published in an unverified section and used nowhere.
DOI 10.5281/zenodo.21787321 (concept — always resolves to the current version)
Where this stands. The paper is recent and not yet cited. The software was released in August 2026 and has no external adopters I can point to. Both statements will stop being true or they will not; either way this page will say which.
In progress
An instrument paper, in English
A methods paper describing the instrument itself — what it measures, what it deliberately refuses to measure, where its weights come from, and how sensitive its output is to them. It makes no claim of priority for the individual components: the IT productivity paradox, the productivity J-curve and the absorption of slack are all long-established, and the paper says so.