Which finance workflows change first
Finance runs on structured documents, which is exactly what generative AI handles best. The first workflows changing are report drafting, earnings and research summarization, first-pass document review in KYC and loan files, reconciliation exception write-ups, and client communication drafts. Analysts who once spent days assembling a briefing pack can now start from a machine draft. Analyses of AI exposure consistently rank finance work as highly exposed at the task level precisely because so much of it is text and rules.
What does not compress is the part regulators care about. Numbers in a filing must be right, a credit decision must be defensible, and a trade must comply with rules that carry personal liability. Generative models are known to produce plausible but wrong figures, which is a serious failure mode in finance specifically. So the realistic near-term change is faster drafts with heavier human checking, and firms building formal review steps around AI output rather than removing people from the chain.
Where human accountability concentrates in finance
Finance is one of the few industries where accountability is written into law with names attached. Executives certify financial statements, auditors sign opinions, compliance officers attest to controls, and advisers owe duties to clients. None of that transfers to a model. As drafting speeds up, the signature points become the bottleneck and the source of value: the person who can look at an AI-assembled analysis and say what is missing, what is wrong, and what they are willing to stand behind.
Judgment also concentrates in ambiguity: pricing a risk with thin data, reading a counterparty in a negotiation, deciding whether an anomaly is fraud or noise, and advising a client whose real question is emotional as much as financial. The ILO's transformation framing fits here: tasks inside finance jobs are being redistributed, with routine assembly moving to machines and interpretation, escalation, and relationship work moving toward people.
What finance workers can build in 90 days
Spend the first month learning your firm's approved AI tools and, just as important, its restrictions, since finance firms often limit what data can enter external models. Practice on low-stakes internal work: meeting summaries, first drafts of routine commentary. In the second month, build a numbers-verification discipline. Every figure in an AI draft gets traced to a source system before it moves forward. Keep a log of the errors you catch; it will teach you the tool's failure patterns.
In the third month, move up the value chain deliberately. Pick one recurring deliverable, use AI to compress the assembly work, and reinvest the saved time in interpretation: what the numbers imply, what you would recommend, and why. In parallel, strengthen regulation and client-facing skills, because exception handling and advice are where analyses suggest finance work is consolidating. Document the before and after; a measured improvement is more persuasive in any career conversation than a list of tools.
