CORAA
Blog/AI in Audit

How to Use ChatGPT and Claude to Calculate Deferred Tax: A Step-by-Step Guide for CAs

Step-by-step prompts for using ChatGPT, Claude or Gemini to build a deferred tax working (DTA / DTL) under AS 22 or Ind AS 12, with a fully worked example, disclosure table and checks for where AI goes wrong.

CCORAA Team26 September 202610 min read

How to Use ChatGPT and Claude to Calculate Deferred Tax: A Step-by-Step Guide for CAs

ChatGPT, Claude or Gemini can build a deferred tax working if you give it the book and tax bases, the applicable framework (AS 22 or Ind AS 12) and the tax rate, and ask it to show every line. It will make arithmetic and recognition errors unless you check them, so use it for the layout and first pass, and verify each figure yourself.

This guide walks through five prompts and a worked example with figures computed by hand, so you can test any AI answer against a known result.

What can ChatGPT or Claude do here, and what should you never delegate?

Reasonable to use AI for Never delegate
Laying out the working table Deciding the tax rate that applies to your client
Classifying items as taxable or deductible differences, for you to confirm Judging whether future taxable profit is probable, or virtual certainty exists
First-pass arithmetic (to be recomputed) Final figures posted to the books
Drafting the note disclosure format Concluding on recoverability of losses or MAT credit

Step 1: Set up a safe workspace

Do not paste client names, PAN, GSTIN, CIN or balances that identify the client. Round the numbers, rename the entity "Company A" and keep the real ledger in your own file. Use a tool setting that does not train on your inputs. See the DPDP-safe prompt template library for anonymisation patterns.

Step 2: Gather the inputs in a table

Deferred tax needs, for each item, the book carrying amount and the tax base at both dates. Collect:

  • Fixed assets: book WDV and tax WDV (see how Schedule II and income tax depreciation diverge).
  • Provisions and accruals allowed only on payment, such as gratuity, leave encashment and bonus items under section 43B.
  • Carried-forward business losses and unabsorbed depreciation.
  • The tax rate and the framework.

Example input (Rs. lakh, synthetic):

Item Opening book Opening tax base Closing book Closing tax base
Fixed assets (WDV) 700 580 800 600
Provision for gratuity 30 0 40 0
Provision for leave encashment 10 0 20 0
Unabsorbed business loss n/a 50 n/a 50

Rate: 25.17% (company under section 115BAA; 22% plus 10% surcharge plus 4% cess = 25.168%).

Step 3: Prompt the AI to classify each item

Act as a chartered accountant's assistant. Framework: [Ind AS 12 / AS 22].
For each item in the table below, state (a) the temporary difference
(book carrying amount minus tax base for assets; tax base minus carrying
amount for liabilities), (b) whether it is a taxable or deductible
temporary difference, and (c) whether it gives rise to a DTL or DTA.
Show the working. If the item is a carried-forward loss, say which
recognition test applies under the framework and do NOT assume it is met.
Do not compute tax yet. List your assumptions.
[paste table]

Expected classification: fixed assets are a taxable difference (book WDV above tax WDV), so a DTL; the two provisions are deductible differences, so DTAs; the loss is a deductible item subject to a recognition test.

Step 4: Prompt the AI to compute DTA / DTL and the movement

Using the classification above and a tax rate of 25.17% (enacted rate
for the reversal period), compute for OPENING and CLOSING:
1. Deferred tax liability and asset per item, rounded to 2 decimals in Rs. lakh.
2. Net deferred tax position.
3. The movement, i.e. the deferred tax charge / (credit) to profit or loss,
   assuming no items go through OCI or equity.
Show each multiplication. Recognise a DTA on the loss ONLY if I state
[recognition evidence: ____]. Otherwise show it as unrecognised.
Re-check the arithmetic and list any assumption.

Step 5: Ask for the disclosure table

Prepare a table for the deferred tax note: for each component,
opening balance, charge/(credit) to P&L, charge/(credit) to OCI, closing
balance, split into deferred tax liabilities and deferred tax assets, then
the net. State whether DTA and DTL can be set off (same taxation
authority and legal right to set off current tax) and list the unrecognised
deductible differences or losses separately. Use only my figures.

What does a good answer look like?

Here is the correct result, computed by hand at 25.17%. AI output will vary from run to run, so treat any AI table as illustrative and compare it with these figures.

Temporary differences

Item Closing difference (Rs. lakh) Type
Fixed assets: 800 - 600 200 (opening 120) Taxable, DTL
Gratuity: 40 - 0 40 (opening 30) Deductible, DTA
Leave encashment: 20 - 0 20 (opening 10) Deductible, DTA

Deferred tax

Item Opening Closing Movement to P&L
DTL, fixed assets 30.20 (120 x 25.17%) 50.34 (200 x 25.17%) 20.14 charge
DTA, gratuity 7.55 10.07 2.52 credit
DTA, leave encashment 2.52 5.03 2.52 credit
Net DTL 20.14 35.24 15.10 charge

Check: the net taxable difference movement is 80 (assets) less 10 (gratuity) less 10 (leave) = 60, and 60 x 25.17% = 15.102, matching the 15.10 charge. Small 0.01 differences in the opening column arise from rounding; the unrounded net opening is 20.136 and closing 35.238.

Loss. The unabsorbed loss of 50 would give a potential DTA of 12.59 (50 x 25.17%). Without the evidence the framework demands (below), it is not recognised and is disclosed as an unrecognised item. If it were recognised at both dates, opening and closing net DTL would each fall by 12.59 (to 7.55 and 22.65) and the P&L charge would stay at 15.10. Recognising it only at closing would instead give a credit of 12.59 in the year.

AS 22 vs Ind AS 12: what changes in this working?

Point AS 22 Ind AS 12
Basis Timing differences between accounting income and taxable income Temporary differences between carrying amount and tax base (balance sheet method)
DTA on ordinary deductible items Recognised where there is reasonable certainty of future taxable income Recognised where it is probable that taxable profit will be available
DTA on carried-forward losses / unabsorbed depreciation Only where there is virtual certainty supported by convincing evidence Only to the extent of probable taxable profit; a history of recent losses requires convincing other evidence
Tax rate Enacted or substantively enacted at the balance sheet date Same
Presentation Set-off only where legal right and same authority Same

In this example the fixed asset and provision numbers are identical under both; the frameworks diverge on losses and recognition tests, not on the arithmetic. Verify paragraph-level wording against the ICAI text for your engagement.

Where does AI go wrong?

  1. Arithmetic. Models mis-multiply or round early. Recompute each line: difference times rate.
  2. Wrong rate. It may pick 30% or 22% by default. For a company under section 115BAA the effective rate is 25.17%; for others use the rate enacted or substantively enacted for the reversal period.
  3. Sign errors. Book WDV above tax WDV is a DTL. Flipped classification reverses the answer.
  4. Mixing AS 22 and Ind AS 12. Watch for "temporary difference" used in an AS 22 answer, or "virtual certainty" in an Ind AS 12 one.
  5. Ignoring recognition tests. AI often books a DTA on losses automatically. Recoverability is your judgment.

Verify in 60 seconds

Enter the same synthetic figures into CORAA's free Deferred Tax Calculator at 25.17%. The net position should match your hand figure of 35.24 net DTL. Then compare your layout with the deferred tax working paper in the Working Papers hub. A mismatch usually means a sign error or a wrong rate.

Which assistant should you use?

ChatGPT, Claude and Gemini can all handle this if you paste in the figures and ask for shown working. Ask any of them to recompute independently, and prefer a tool that can run a spreadsheet or code for arithmetic where available. None replaces your own check.

Last reviewed: 26 September 2026.

Need to run deferred tax working papers across clients? Start a CORAA trial.

Frequently asked questions

Can ChatGPT calculate deferred tax accurately?

It can lay out the working, but arithmetic and rate errors are common. Recompute each line and verify against a calculator.

What tax rate should I use for deferred tax in India?

The rate enacted or substantively enacted at the reporting date that applies when the difference reverses. For a company that has opted for section 115BAA, the effective rate is 25.17%.

Do I recognise a DTA on carried-forward losses?

Only if the framework's test is met: virtual certainty with convincing evidence under AS 22, or probable taxable profit with convincing evidence where there is a loss history under Ind AS 12.

Is deferred tax on fixed assets always a DTL?

No. It is a DTL when book WDV exceeds tax WDV, and a DTA when tax WDV exceeds book WDV.

Can DTA and DTL be netted?

Only where the entity has a legally enforceable right to set off current tax and the balances relate to the same taxation authority.

Should I paste the client trial balance into the AI?

No. Use anonymised, rounded figures and keep real data in your own files.

Topics
how to calculate deferred tax using ChatGPTdeferred tax calculatordeferred tax calculation Ind AS 12deferred tax AS 22 exampleDTA DTL calculationClaude for deferred taxAI for CA deferred tax working
Share
← Back to all articles
Keep reading

More in ai in audit.

Built for India · DPDPA compliant

Ready to automate your audit work.

See how Coraa reduces audit engagement time by 60%, from ledger scrutiny to working papers, all from one Tally import.

Run one complete audit free