AI Tools for Chartered Accountants in India: What ICAI Has Actually Agreed

ICAI’s tie-up with ISCA is mainly about separating a proposed AI fluency programme from live qualification changes: the ai tools for chartered accountants india angle remains exploratory, while specified SCAQ exam exemptions for eligible ICAI members took effect on 21 August 2026, subject to remaining entry, assessment and qualification requirements.

The short answer

On 21 August 2026 ICAI and the Institute of Singapore Chartered Accountants

announced two separate things, and the difference between them matters. One is a

proposed AI training programme that does not exist yet. The other is a set of exam

exemptions that took effect the same day. If you are a CA wondering what changed

this week, the exemptions are the part you can act on.

What was announced, precisely

The AI programme is proposed, not available

ISCA signed a Letter of Cooperation setting out a framework for a proposed **Global

AI Fluency Initiative**, centred on ISCA’s existing AI Fluency Programme. The wording

in the release is careful: the parties “will explore how ISCA and ICAI can work

together to extend practical and responsible AI learning to ICAI members and

students, subject to relevant approvals and mutually agreed implementation

arrangements.”

Read that as an agreement to explore, not a course you can enrol in.

The programme it is based on runs about 30 hours and comprises **180 hands-on

cases based on workplace applications**. If extended, it would reach ICAI’s community

of more than 1.5 million members and students. The release describes the content

as structured around AI fluency foundations, programme promotion between the

institutes, member outreach, and professional application across accounting, audit,

finance and tax.

*(Worth noting for accuracy: the release says “three areas” and then lists four

bullets, two of which describe promotional arrangements rather than course content.

The substantive learning areas appear to be AI fluency foundations and professional

application.)*

Mr Lee Boon Teck, President of ISCA, said AI “is reshaping the accountancy profession

and building AI fluency alongside strong professional expertise will be critical to

enabling the profession to adapt and create greater value.”

The SCAQ exemptions are already in force

This is the concrete part. ISCA approved a proposed Module Exemption Agreement

with ICAI covering both bodies’ qualification pathways. The agreement itself has not

been formally executed. Even so, the release states plainly that **ISCA will

implement the specified exemptions from the SCAQ for eligible ICAI members with

effect from 21 August 2026**.

Eligible candidates must still meet the applicable entry, assessment and remaining

qualification requirements. The exemptions reduce duplication; they do not hand

anybody a qualification.

Mr Prasanna Kumar D, President of ICAI, described the exchange of letters as “a

defining step in our shared journey and a clear statement of intent to work towards

strengthening recognition between our professional qualifications.”

For context on the counterparty: ISCA is Singapore’s national accountancy body,

established in 1963, designated by Singapore’s Ministry of Finance to confer the

CA (Singapore) designation. It supports over 46,000 members across more than 40

countries.

So where do AI tools actually fit in an Indian practice?

The announcement says nothing about which software to buy. That question is still

yours to answer, and the sensible way to think about it is by how much professional

judgement a task carries. Adopt from the bottom of that list upward.

Document extraction and data entry

This is the highest-volume, lowest-judgement work in most practices. Pulling figures

off bank statements, purchase invoices, Form 16s and broker contract notes into a

usable sheet. Language models combined with OCR handle messy layouts far better than

the template-based tools that came before.

The catch is that extraction confidence is not uniform. A tool that reads 200 invoices

correctly can misread the 201st, and misread it plausibly, producing a number that

looks entirely reasonable. Any workflow here needs reconciliation against a control

total rather than a spot check.

Drafting and summarising

First drafts of engagement letters, management representation letters, board minutes

and client explainer emails. Also condensing long documents, such as a lease or a

partnership deed, into an issues list you then verify yourself.

Remember what a summary is: a lossy compression chosen by a model. It will drop the

clause that mattered as readily as the one that did not.

Research and first-pass interpretation

Finding the relevant section, circular or ruling, and getting a plain-language

explanation to orient yourself before reading the real thing.

This is where the cost of error is highest. Language models routinely produce

citations that do not exist, in correct format, with plausible numbers attached.

Every section, circular number, case name and date needs checking against the primary

source, whether that is incometaxindia.gov.in, the CBDT circular itself, or the

court’s own order. Treat the model as a pointer to where to look. Never as the

authority on what it says.

Reconciliation and anomaly detection

GST 2A and 2B against purchase registers. Form 26AS and AIS against books.

Inter-company balances. Pattern-spotting across large ledgers to flag what deserves a

human look.

Anomaly detection produces false positives and false negatives alike. It narrows

where you look. It does not certify what it failed to flag.

What does not delegate

Opinion. Sign-off. Materiality judgement. Professional scepticism. The decision about

what a client actually needs. These carry your membership number, and ISCA’s own

framing of the AI programme puts professional judgement and oversight at the centre

rather than at the edges.

The confidentiality question nobody should skip

Uploading a client’s books, PAN-level data or draft financials to a consumer AI

service sends confidential information to a third party, quite possibly outside

India, on terms you did not negotiate.

ICAI’s Code of Ethics requires members to maintain the confidentiality of information

acquired through professional and business relationships. Disclosure is permitted in

defined circumstances: where required by law; where permitted by law and authorised

by the client; and where there is a professional duty or right to disclose, such as

complying with Peer Review or Quality Review requirements, responding to a regulatory

inquiry, or protecting professional interests in legal proceedings. Disclosing

confidential information acquired in the course of employment, other than as required

by law, is treated as professional misconduct.

None of those exceptions obviously covers “I pasted it into a chatbot to save time.”

Before any tool touches client data, settle four things in writing: where data is

processed and stored, whether inputs are used to train the provider’s models,

retention and deletion terms, and whether your engagement letter permits third-party

processing at all.

Reviewer note: the confidentiality provisions above are summarised from ICAI’s

published Code of Ethics. Before publishing, cite the specific clause numbers from

the current edition on icai.org. I have deliberately not guessed at them.

A practical adoption sequence

• Start where a mistake is visible. Data extraction with a control-total check fails

loudly, which is exactly what you want from a first deployment.

• Never let the tool be the last step. Whatever it produces gets read by a person who

is accountable for it.

• Decide the data rule before choosing the tool. Which client data may leave the

firm, to which provider, on what terms. Written down once, for everyone.

• Verify every citation against the primary source, however well-formed it looks.
• Keep a record of your own checks. If a review or a dispute follows, “the software

produced it” is not a defence.

Frequently asked questions

Can I enrol in the ICAI AI programme now?

Not yet. As announced on 21 August 2026, ISCA signed a Letter of Cooperation for a

proposed Global AI Fluency Initiative, explicitly subject to approvals and mutually

agreed implementation arrangements. Watch icai.org for enrolment details.

What are the SCAQ exemptions, and are they real?

They are real and already operating. ISCA implements specified exemptions from the

Singapore Chartered Accountant Qualification for eligible ICAI members with effect

from 21 August 2026, even though the Module Exemption Agreement has not yet been

formally executed. Eligible candidates must still meet applicable entry, assessment

and remaining qualification requirements.

How long is the AI programme, if it goes ahead?

Approximately 30 hours, built around 180 hands-on cases based on workplace

applications.

Can I use AI to prepare a client’s return?

Tools can assist with extraction, reconciliation and drafting. The return, the

positions taken in it and the signature remain your professional responsibility, and

every figure and legal position needs verifying against the primary source.

What is the single biggest risk?

Fabricated citations. A model will give you a section number and a confident summary

of what it says, and be wrong about both. Check every one.

Sources

• “ISCA and ICAI Deepen Collaboration on AI Fluency and Professional Recognition”,

Media OutReach Newswire, Singapore, 21 August 2026. Verified verbatim against two

independent mirrors:

https://thearabianpost.com/isca-and-icai-deepen-collaboration-on-ai-fluency-and-professional-recognition/

and https://macaubusiness.com/isca-and-icai-deepen-collaboration-on-ai-fluency-and-professional-recognition/

• “ISCA, ICAI partner on responsible AI learning for accountants”, *The Accountant

Online* (GlobalData), Ellichipuram Umesh, 24 August 2026.

• ICAI Code of Ethics, published at icai.org and kb.icai.org. **Reviewer: cite clause

numbers from the current edition before publishing.**