What the evidence says
- Adoption is broad: Gartner found 59% of finance functions using AI in 2025 (58% in 2024, 37% in 2023). The top uses were knowledge management (49%), accounts payable automation (37%) and error/anomaly detection (34%).
- Impact is lagging: in the same survey, 91% reported low or moderate impact. In Gartner’s March 2026 survey, 63% of finance leaders said implementation in 2025 was slower than expected.
- Leaders pull ahead: KPMG’s global study of 2,900 organisations (Dec 2024) found 71% using AI in finance, and the most advanced users far more likely to say returns beat expectations.
The lesson: pick narrow, high-volume workflows, measure the hours saved, and design review steps in from day one.
Where to start
| Workflow | What AI does | What a person still does |
|---|---|---|
| Accounts payable | Reads invoices, suggests ledger and GST coding, matches to POs | Approves coding exceptions and payments |
| Bank & gateway reconciliations | Matches transactions, groups exceptions | Investigates and clears exceptions |
| GST input tax credit | Compares purchase register with supplier data before GSTR-2B | Decides accept / reject / pending in the Invoice Management System |
| Month-end close | Drafts accrual schedules and flux commentary | Reviews and posts journals |
| FP&A | Runs scenarios, explains variances in plain English | Sets assumptions, owns the forecast |
| Investor updates | Drafts the monthly update from the MIS | Checks every number and the narrative |
A 90-day rollout
- Weeks 1–2 — map and measure. List recurring finance tasks with hours per month. Pick two or three with high volume and clear rules.
- Weeks 3–4 — choose tools. Prefer features inside the accounting and banking tools you already use; add a general AI assistant on a business plan that excludes your data from training.
- Weeks 5–8 — pilot in parallel. Run the AI workflow alongside the manual one for a full close. Compare accuracy and time.
- Weeks 9–12 — switch over with controls. Write down who reviews what, keep prompts and outputs, and post everything through the ledger.
Indian rules to design around
- Audit trail (Companies Act): since 1 April 2023, companies must keep books in software that records an edit log of every change, which cannot be disabled, with daily backups on servers in India. Auditors report on it. AI-prepared entries should go through that software — never be pasted in as untracked adjustments.
- DPDP Act and Rules 2025: the rules were notified in November 2025 and phase in. The core duties — notice and consent, security safeguards, breach reporting and retention limits — apply from 13 May 2027 under the notified timeline. Your company stays responsible for personal data it gives an AI vendor, which should act under a contract.
- RBI FREE-AI (for banks, NBFCs and other regulated entities): RBI’s August 2025 framework sets 7 principles (“sutras”) and 26 recommendations, including a board-approved AI policy, lifecycle governance, AI-specific incident reporting and telling customers when they deal with AI. In June 2026 RBI also issued draft guidance on model risk management covering AI/ML models.
- SEBI (for brokers, advisers, AMCs and other intermediaries): since February 2025, a SEBI-regulated entity using AI is solely responsible for the data it uses and for the outputs. Broader guidelines on responsible AI use were consulted on in June 2025 — check sebi.gov.in for the final circular.
- MeitY: India’s AI Governance Guidelines (November 2025) are voluntary and principle-based; there is no standalone AI law.
If you operate abroad
- EU: the EU AI Act’s high-risk rules for credit scoring and life/health insurance pricing now apply from 2 December 2027, after the July 2026 “digital omnibus” amendment. Transparency duties apply from 2 August 2026.
- UAE: no comprehensive AI statute; the Central Bank’s February 2026 guidance note sets expectations for licensed financial institutions.
- US: rules are state-by-state; Colorado’s re-enacted AI law takes effect on 1 January 2027.
Common mistakes
- Letting AI output reach investors or tax filings without a named reviewer.
- Pasting payroll or customer data into consumer AI accounts.
- Posting AI-prepared journals outside the accounting system, which breaks the audit trail.
- Automating a messy process instead of fixing the chart of accounts and close checklist first.
- Measuring “AI usage” instead of hours saved and errors caught.
Want help putting this in place? See AI in finance advisory or the Virtual CFO service.
We also help businesses get found in AI answers — AEO services (answer engine optimization).
Frequently asked questions
What percentage of finance teams use AI?
Gartner's 2025 survey of finance leaders found 59% of finance functions using AI, up from 37% in 2023; KPMG's December 2024 global study put the figure at 71%.
What is RBI's FREE-AI framework?
The Framework for Responsible and Ethical Enablement of AI, released by RBI on 13 August 2025, sets out 7 principles and 26 recommendations for regulated entities, including a board-approved AI policy, governance across the model lifecycle, AI incident reporting and customer disclosure.
When do the DPDP Rules apply to companies?
The DPDP Rules 2025 were notified in November 2025 and phase in; the core obligations for data fiduciaries, such as notice, consent, security safeguards and breach reporting, apply from 13 May 2027 under the notified timeline.
Can AI-generated journal entries be posted in Indian companies?
Yes, provided they are posted through accounting software that maintains the mandatory audit trail (edit log) and are reviewed and approved by an accountable person.
Please note: this guide is general information based on the rules as they stood on the review date above. Regulations and RBI/CBDT procedures change often — please get advice on your specific facts before acting.