Blog— Anton Ignashev

JPK_ST_KR in an accounting office: the draft exemption to 2028 covers sending the file, not the data

JPK_ST_KR in an accounting office: the draft exemption to 2028 covers sending the file, not the data

The Exemption Everyone Heard About Is About Sending, Not Keeping

In August 2026 the Ministry of Finance published a draft regulation that would exempt CIT taxpayers and partnerships that aren't legal persons from sending JPK_ST_KR, the fixed-asset part of JPK_CIT, for tax years starting in 2026, 2027 and 2028. The headlines invited an obvious reading: "so fixed assets are off the table until 2029".

They aren't. There are three reasons. The draft is about sending the file, not about what the register has to contain. It doesn't cover every client you have. And most of the data the structure asks for is something the tax office can request in an audit anyway — and something your JPK_KR_PD for 2026 already depends on.

First, a note on status. As I write this, on 5 October, the regulation is a draft, and drafts change during the legislative process. Check whether it has been signed and published before you build a plan around it. The advice below holds either way.

Who the Draft Doesn't Cover

The draft, as the advisory firms describe it, covers CIT taxpayers and partnerships that aren't legal persons. None of the commentary I've read extends it to PIT taxpayers on full books.

For a typical accounting office, that's not a rounding error. Sole traders and civil partnerships that crossed the full-books threshold can make up a real share of the full-books portfolio. They're also the clients whose fixed-asset registers are most likely to be in poor shape. A few years ago many of them were on the revenue and expense ledger (KPiR), the register was a spreadsheet, and nobody ever split book depreciation from tax depreciation. Nobody had to.

So the first job this month is a list: every full-books client, marked CIT or PIT. For the PIT ones, assume the fixed-asset data for 2026 is reportable until a published regulation tells you otherwise.

Even for CIT clients, the commentary doesn't agree on what the draft lifts. Deloitte reads it as leaving untouched the date from which the register must meet the requirements — 1 January 2025. Others read it more broadly. When the experts disagree, the cheap move is to keep the data in shape and treat the exemption as relief on sending, not permission to stop.

What the Structure Actually Asks For

JPK_ST_KR describes each fixed asset and intangible in fields E_1 to E_32. The inventory number and acquisition date won't cost you much time. The fields that will are the ones that assume two separate views of the same asset:

  • A tax depreciation rate and a book depreciation rate. E_10A and E_10B carry the tax rate as a percentage and as an amount (the amount is where one-off depreciation lands); E_25A and E_25B carry the book equivalents.
  • Tax and book initial values, and their changes. Improvements, revaluations, partial disposals.
  • Accumulated depreciation in both views, for the year and in total.
  • A KŚT symbol for every fixed asset (intangibles don't have one).
  • Disposal data where an asset was sold or scrapped, including the reason.

In most small companies, book and tax depreciation are the same number. That's a legitimate policy choice, not an error. The trouble is that "the same" is usually a default nobody actually chose. Someone set up the register software once, copied the book rate from the tax rate, and nobody looked at the two columns separately again. The structure won't let you report one column and leave the other blank.

The tax authority has also shown how it reads "not applicable". In an interpretation of 25 May 2026 (0111-KDIB2-1.4010.80.2026.1.ED), a company on Estonian CIT asked whether it could put zeros in the tax fields, since it no longer depreciates for tax purposes. For assets acquired before it switched regimes, the answer was no: the historical tax value and past tax depreciation had to go in. Zeros were accepted only for assets bought after the switch.

The lesson for an office? If you can't reconstruct the tax history of an old asset today, you have a problem — whenever the file turns out to be due.

Four Places the Register Is Usually Wrong

These are the patterns I'd look for first, roughly in the order they show up.

1. The register doesn't reconcile with the ledger. The gross value of fixed assets in the register should equal the balance of account 010 (or its analytics), and accumulated depreciation should equal account 070. In practice, the register lives in a separate module or a spreadsheet, and the ledger is posted from a monthly depreciation summary. Give it a few years and a gap opens: disposals that never reached the register, assets posted to 010 without a register card, manual corrections. This is the cheapest check to run and the most useful, because a gap here means the file and the books will contradict each other.

2. KŚT symbols are missing, old or don't match the rate. Assets entered before the current classification (KŚT 2016) often carry an old symbol, a group-level symbol, or nothing at all. Then there's the other version of the same problem: a symbol that implies one statutory rate sitting next to a different rate in the register, with no individual rate decision or increase coefficient to explain it.

3. Repairs that should have been improvements. Under the CIT Act, spending on rebuilding, extending or modernising an asset increases its initial value when it exceeds 10,000 zł in a tax year. In practice, large invoices for "remont" go straight to costs, because they were booked from the description rather than from what was actually done. Grant Thornton names repair-vs-improvement as one of the areas JPK_CIT makes visible, and I agree. It's also the hardest to fix after the fact: change an initial value and every depreciation charge since then changes with it.

4. Low-value assets handled inconsistently. Assets up to 10,000 zł can be expensed directly. If the taxpayer chose that route, they belong in neither the register nor the file. The problem is the mix — some low-value items on register cards and depreciated, others expensed, with no consistent rule per client. Any rule will do. No rule means questions.

What It Costs Across a Portfolio

Rough arithmetic again, same office as in my JPK_CIT post: 40 full-books clients. Say an average of 30 register cards per client with movement or a balance, so 1,200 assets across the portfolio.

A full manual review at five minutes per card (symbol, both rates, both values, history) is 100 hours. Reconciling the register to accounts 010 and 070 takes maybe 30 minutes per client when the figures agree, and anything from two hours to a day when they don't. If one client in four has a gap, that's another 40–60 hours.

Most of that time is not judgement. It's opening cards and comparing numbers.

What an Agent Does, and What Stays With the Accountant

It's very close to the work I described for chart-of-accounts markers. Lots of reading, a short list at the end.

The agent can, read-only:

  • reconcile the register's gross value and accumulated depreciation with accounts 010 and 070 for every client and report the gap per client, per month;
  • list cards with no KŚT symbol, an outdated symbol, or a tax rate that doesn't match the statutory rate for the symbol;
  • list cards where book and tax rates and values are identical, so the accountant can confirm that's a policy and not a default;
  • scan cost accounts for repair and modernisation invoices above 10,000 zł per asset per year and put them next to the asset cards they probably belong to;
  • check low-value items for a consistent rule per client: register card or direct expense, not both.

The accountant keeps:

  • every decision on classification, rates and whether spending was a repair or an improvement. Those are tax decisions;
  • any correction of initial values and the re-calculation that follows;
  • the conversation with the client about what was found.

It's the same split as in every project I do in accounting: the agent prepares, a person signs. I wrote about where that line sits in AI agents and liability.

A Calendar That Doesn't Depend on the Draft

  • October: list clients by stage and by tax (CIT vs PIT on full books). Reconcile register to 010 and 070 for each. Count cards with missing or mismatched KŚT symbols.
  • November: fix the register for PIT clients first, since the draft doesn't cover them. Review repairs above the threshold. Do it alongside the year-end balance confirmations and the chart-of-accounts work for JPK_KR_PD.
  • December: check whether the regulation was signed. Either way, the register is in shape before the year-end close starts.

One Number to Start With

If you do one thing this month, take your five largest full-books clients and compare the register's gross value with the balance of account 010 at 30 September. Count how many agree to the grosz.

Five out of five? Your registers are in better shape than most, and the exemption really does buy you time. Two? Then the exemption only postpones finding out. If you'd like a second pair of eyes on that comparison, or want to see what it looks like when an agent runs it across the whole portfolio, write to me.

Check a Polish VAT counterparty in seconds, free

KsięgoAI checks a NIP against the VAT white list and VIES, scores the risk and saves proof of due diligence. Works in Telegram and Claude.

Try KsięgoAI→

Let’s talk about your project

Free 30-minute consultation. We’ll figure out if and how I can help.

Book a Free 30-Minute Call

Select a date

November 2026
Mon
Tue
Wed
Thu
Fri
Sat
Sun
← Back to Blog

Related Posts

What Subiekt nexo and Subiekt GT cost in 2026: licence prices, the Abonament, and the 30 days that raise the renewal by half
Blog

What Subiekt nexo and Subiekt GT cost in 2026: licence prices, the Abonament, and the 30 days that raise the renewal by half

Subiekt nexo for 3 workstations is 1 275 zł net, Subiekt GT 1 045 zł, nexo PRO 2 575 zł. Those are perpetual licences, and then there's the yearly Abonament: 349 zł for nexo if you renew within 30 days of expiry, 539 zł if later. Plus Sfera (the API), the accounting-office client licences, and a comparison with Optima and enova365.

Read more→
What Comarch ERP Optima costs in 2026: the subscription list, Basic vs Pro, and three costs it leaves out
Blog

What Comarch ERP Optima costs in 2026: the subscription list, Basic vs Pro, and three costs it leaves out

Optima's price list is public: Podstawowy 59 zł a month, Logistyka Basic 75 zł, Księga Handlowa 175 zł. The catch is the word Pro: it's a top-up on Basic, not a separate tier, so full accounting in Finanse costs 250 zł, not 200. Plus the Podstawowy limits, how HR counts employees for accounting offices, and two worked examples.

Read more→
December pay paid in January 2027: a 2026 or a 2027 cost? The 8 January payday and ZUS by 15 or 22 February
Blog

December pay paid in January 2027: a 2026 or a 2027 cost? The 8 January payday and ZUS by 15 or 22 February

December payroll paid in January is a 2026 cost, but only if the pay leaves on the date in the client's pay rules and the ZUS contributions reach ZUS on time. 10 January 2027 is a Sunday, so the statutory deadline ends on Friday 8 January. The same pay lands in the employee's 2027 PIT-11. Here's what an office checks in December and in February.

Read more→