Most AI in finance is solving the wrong problem.
Lucy is solving the right one.
Doctoral researcher · National media commentator
Nobody understands micro-business financial data the way Lucy Cohen does. After two decades processing the transactions of thousands of small businesses through Mazuma, she has seen the patterns, the exceptions, the failures and the fixes at a scale very few researchers can claim.
That practitioner knowledge became the foundation of an active doctoral research programme. Lucy's research sits at the intersection of AI in financial settings and the reality of how small businesses actually work: not in theory, but in the messy, high-volume, error-prone reality of day-to-day bookkeeping.
Her findings challenge some of the profession's most widely held assumptions about AI, automation and the role of human oversight. They are inconvenient for some. They are essential for all.
Areas of expertise
Micro-business financial data:
The unique characteristics, failure modes and processing challenges of financial data at the small end of the market: sole traders, micro-businesses and the firms that serve them.
AI in financial settings:
How artificial intelligence is being applied, misapplied and misunderstood in accounting and bookkeeping: what works, what doesn't, and why the difference matters.
Probabilistic vs deterministic systems:
The case for rethinking which types of AI are appropriate for financial data processing, and the consequences of getting that decision wrong.
Human oversight and AI autonomy:
When human review of AI systems adds value, and when it creates risk. A practitioner and researcher perspective on the oversight debate.
The papers
Lucy's papers are available to read online on Substack. Enter your email on Substack to receive updates when new research is published.
Citation and sharing
All papers are published under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International licence (CC BY-NC-ND 4.0). You are free to share and redistribute them in any medium or format, provided you give appropriate credit to the authors, include a link to the licence, and do not use the material for commercial purposes or distribute adapted versions of it.
To cite these papers, please use the following format:
Cohen, L. (2026). Probably Wrong: Stochastic Contamination and the Case Against Probabilistic AI in Financial Data Processing. Retrieved from lucycohen.uk
Cohen, L. & Lothian, J. (2026). Get Out of the Loop: The Case Against Human Oversight of Deterministic AI Systems. Retrieved from lucycohen.uk
For permissions beyond the scope of this licence, contact
Media and commentary
Lucy is a regular contributor to national media on AI in finance, the future of accountancy and the regulatory environment for small businesses. She has appeared on BBC, Sky News and ITV, and written for Forbes.
For media enquiries and interview requests, contact Lucy directly via the Work with Lucy page.
Want Lucy to speak on AI in finance at your event or contribute to your publication? See her speaking topics on The Stage, or get in touch via Work with Lucy.
