AI in Finance #001 — What Are the Real Barriers?
Originally published in Weekend Reads on LinkedIn · Subscribe
The AI wave isn't slowing down — not in usage or adoption. What's changed is that organizations have realised building with AI is fast, but token costs can quietly overwhelm the value the output actually produces. So the noose is tightening: teams are running hard cost-benefit checks on money spent versus outcome delivered.
But here's what surprised me. I mapped what 28 major players are doing with AI and the challenges slowing them down — and compute cost and infrastructure economics sit near the bottom of the list. The biggest challenge, by a distance, is regulatory and compliance burden.
For this series, I've kept the scope to the BFSI domain. It's no secret that this sector already runs a continuous race against changing regulation — and AI, with the new policy layer forming around it, only adds complexity.
The signals are everywhere. In the EU, the AI Act treats credit scoring and insurance pricing as "high-risk" — and while those obligations were just pushed from August 2026 to December 2027 under the Digital Omnibus, the delay tells its own story: even the world's most prescriptive AI law couldn't ship its finance rules on time. In the UK, Parliament's Treasury Committee has recommended that the FCA to publish practical AI guidance for firms by the end of 2026 — including how senior-manager accountability applies when AI causes customer harm. And in India, SEBI's Regulation 16C already makes a regulated entity responsible for the outputs of any AI tool it deploys, with the RBI's FREE-AI framework pointing the same way.
The applications of AI are enormous in the BFSI domain — automating routine work, running statistical models, trading, risk analysis; almost every function here can be sharpened by AI. But how do you justify your position when a trade goes wrong, or a client sues over a bad recommendation? How do you answer a regulator when a compliance check you handed to an AI agent gets missed?
Read the AI rules in any of these geographies and one thing is constant: the organization is fully responsible for what its AI produces. That's why AI governance, guardrails and harness engineering are the skills worth learning — the way to move at AI velocity without trading away quality or control.
Over the coming weeks I'll break down real cases where companies were legally challenged, what I think the optimal solution would have been, and more on AI in finance.
Keep watching this space. Every Saturday.
References
European Union
— EU AI Act (Regulation 2024/1689), high-risk classification incl. credit scoring & insurance pricing — Annex III: eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32024R1689
— Digital Omnibus (Regulation 2026/1744) deferring high-risk obligations to Dec 2027 — analysis: gibsondunn.com/eu-ai-act-omnibus-agreement-postponed-high-risk-deadlines-and-other-key-changes
United Kingdom
— Treasury Committee, "AI in Financial Services" (HC 684): publications.parliament.uk/pa/cm5901/cmselect/cmtreasy/684/report.html
— FCA, AI in financial services hub: fca.org.uk/firms/ai-financial-services
India
— SEBI (Intermediaries) (Amendment) Regulations 2025 — Regulation 16C: sebi.gov.in/legal/regulations/feb-2025/securities-and-exchange-board-of-india-intermediaries-amendment-regulations-2025_91809.html
All sources verified as of 14 August 2026. Regulatory positions in this area move quickly — check the primary source before relying on any point.
AI in Finance
Have thoughts on this?
Let's continue the conversation on LinkedIn.
Get new posts by email
Occasional notes on the BA → AI PM journey. No spam.