AI in Finance.
What AI actually changes for finance work — what's real, what's hype, and where it already earns its keep.
AI in finance is the practical question of what artificial intelligence actually changes for finance work — and what it doesn’t. Today it lands first on the routine, high-volume paths: transaction coding, reconciliations, variance commentary, and contract and invoice extraction, in a draft-and-review pattern where the model produces the first pass and a human owns the judgment and the sign-off.
What it doesn’t change is accountability — when the model is wrong, the controller still owns the number. Strategy of Finance covers what’s working in production, what’s still vapor, and where AI earns its keep — close speed, FP&A turnaround, spend analysis — alongside the guardrails (the three-eyes principle, exception logging, and reserving model output for cheap-to-verify cases) that keep a confident-but-wrong answer from compounding into an audit finding.
Essays 24
The Credit Index reads: the coding tools Pricing Who still gets to keep 75%+ gross margins? Pricing What you charge for matters more than what you charge Pricing “Unlimited*” — the asterisk is the business model Pricing The price didn't change. Your bill did. Pricing The AI pricing plateau Pricing Should you meter at all? Pricing How to move a meter without a mutiny Pricing How to buy AI without overpaying Pricing Credits are how AI hides its prices Pricing Almost no one can charge for outcomes Pricing AI pricing is just COGS finally showing up Pricing I just graduated again. Thanks, Claude. AI in Finance The software you didn't buy AI in Finance Vulnerability is a grace reserved for the wealthy Talent & Equity AI gives, AI takes: the new physics of non-physical businesses AI in Finance Demo to production — Harsh Joshi on the chasm enterprise AI keeps falling into Podcast Summary Less close, more counsel — David Junius on the strategic CFO Podcast Summary The AI-finance paradox AI in Finance I want to be a CFO — Joy Mbanugo on the deliberate path Podcast Summary Underrated CFO traits for the AI era AI in Finance What I'd tell my pre-AI self AI in Finance What to look for in a SAM tool Procurement / Spend How to evaluate finance software Modern Finance Function
Episodes 2
EP 029 Harsh Joshi DAO Studio EP 029 - Demystifying Artificial Intelligence - AI 101 with Harsh Joshi, Founder & CEO at DAO Studio Jun 11, 2024 1 hr 25 min EP 009 Joy Mbanugo ServiceRocket EP 009 - Confidence, Prepare, Execute, and Repeat with Joy Mbanugo, CFO at ServiceRocket Aug 14, 2023 1 hr 7 min
Related questions
Reviewed- What does AI actually change in a finance function today?
- Mostly the routine, high-volume paths first — transaction coding, reconciliations, variance commentary, and contract and invoice extraction. The pattern is 'draft-and-review': AI produces the first pass, a human owns the judgment and the sign-off. What it doesn't change is accountability — when the model is wrong, the controller still owns the number. The near-term ROI shows up in close speed, FP&A turnaround, and procurement spend analysis, not in replacing the people who interpret the output.
- Will AI replace finance jobs?
- It replaces tasks faster than roles. The most exposed work is the repeatable, rules-based middle — the parts of FP&A, AP/AR, and reporting that are pattern-matching dressed up as judgment. The work that compounds in value is the opposite: judgment under uncertainty, narrative for boards and investors, and knowing which questions to ask. Finance professionals who move up the judgment curve get leverage from AI; those who stay in the rote middle get compressed by it.
- Where should a finance team start with AI?
- Where the work is high-volume, low-ambiguity, and easy to check: reconciliations, spend categorization, contract data extraction, and first-draft variance narratives. Keep a human in the loop, log every exception, and measure the share of volume handled end-to-end without escalation — that number, not the demo, tells you whether it's working. Avoid betting first on the high-stakes, low-verifiability cases (forecasting the number, board guidance) where a confident-but-wrong model does the most damage.
- What's the biggest risk of AI in finance?
- Automating a process you don't understand. AI will happily produce a clean, confident, wrong answer, and finance is exactly the domain where that compounds — into a misstated forecast, a controls gap, or an audit finding. The guardrails are old-fashioned: the three-eyes principle (the parties who do, check, and approve a transaction are different), exception logging, and reserving model output for places where the answer is cheap to verify.