Predictika · ASC 740 uncertain tax positions

An AI tax agent's uncertain tax position, checked before it is recorded

An AI tax agent prepares an uncertain tax position entry under ASC 740 (formerly FIN 48): the authority behind the position, its technical merit, the settlement outcome table and any lapse or settlement. Before it is recorded, the agent sends it to Predictika, which checks it against a declarative model of the recognition and measurement rules and returns a verdict, every reason and the figures it computed.

Scope: ASC 740 uncertain tax position entries: the more-likely-than-not test, measurement, reserve, interest, statute of limitations and settlements. See everything this validator covers

This validator checks proposed uncertain tax position entries under the ASC 740-10 recognition and measurement approach — not just the scenarios shown below.

  • Recognition — a position is recognized only when its technical merit is above 50 percent; exactly 50 percent is not recognized.
  • Measurement — the benefit is measured as the largest outcome in a three-outcome table with a cumulative probability of at least 50 percent; the last outcome must be at 100 percent; the reserve is the gross benefit less the measured benefit.
  • Interest and limitations — interest must be accrued on the unrecognized benefit, and the reserve must be reversed once the statute of limitations has expired.
  • Settlements — an effective settlement must equal the measured benefit.
  • Review items — recognition that rests on no direct authority (as against a statute, regulation, court case or private letter ruling) goes to a partner, and a reserve above the materiality threshold goes to disclosure review.

Any uncertain tax position entry under these rules is in scope, not only the scenarios pictured here. Not covered: deferred taxes and valuation allowances, penalties, state and foreign rules beyond a jurisdiction label, tax rates, discounting, interim reporting and the technical analysis itself. The model is a simplified illustration of public ASC 740 guidance with the thresholds listed above; it is not accounting advice.

The AI tax agentPrepares the uncertain tax position entry from the tax workpapers and says it is ready to record.
Predictika validatorChecks every rule at once and returns the verdict, each finding and the computed figures.

7 scenarios in 3 groups · each group runs from the agent’s first attempt to the outcome · click one to switch

What the agent proposes

The AI tax agent says

Illustrative agent statement. The validator receives the structured request below.

Proposing agent: AI tax agent (scripted)

Before recording, the agent must call Recorded run

Predictika validator


      
Waiting for the call.

Where typical rules-based tools get it wrong

Click to expand

Our assessment of what a typical rules-based tool (validation rules, workflow checks and approval routing configured in an ERP, tax-provision or governance system) does with the same cases. It is our assessment, a judgment about effort and reliability, not a claim that nothing can be built.

What the validator caughtTypical rules-based toolPredictika
The largest outcome that is at least 50 percent likelyOnly if someone is vigilant
A lookup across a three-outcome table with nested conditions. It works, but boundaries such as exactly 50 percent are easy to get wrong.
Computed from the table on every call; the 50 percent boundary is tested on both sides.
Interest and reversal tied to a calculated reserveOnly if someone is vigilant
Each check depends on a reserve that has to be calculated first, so the order of the rules matters.
The reserve is computed first, then every dependent check runs on the same call.
A settlement amount that must equal a calculated measured benefitOnly with workarounds
It compares an entered amount with the result of a multi-step calculation, so it only fires after the calculation is complete and saved.
Compared with the measured benefit on the same call as the calculation.
Every problem returned in one answerOnly with workarounds
Checks tend to fire one at a time as steps are saved, so a preparer fixes one error and meets the next.
All findings come back together, whatever order the choices arrived in.

What this validator catches

Click to expand

Every rule is checked on every call, in any order the choices arrive; one problem never hides another. Blocking findings stop the request; approval findings need a person to sign off; everything else is computed.

1A position is recognized only when its technical merit is above 50 percent; exactly 50 percent is not recognized.
2The measured benefit is the largest outcome with a cumulative probability of at least 50 percent.
3The reserve for the unrecognized portion is the gross benefit less the measured benefit.
4Interest must be accrued on any unrecognized benefit.
5A position whose statute of limitations has expired must have its reserve reversed.
6An effective settlement must equal the measured benefit.
7The cumulative probability of the last outcome must be 100 percent.
8Recognition that rests on no direct authority needs partner review.
9A reserve above the materiality threshold needs disclosure review.

Scenarios and results

Click to expand

The recorded result of the validator for each scenario on this page: the request the AI agent sent, the verdict, and every finding, in the order they appear above.

1. Lapsed statute of limitations: A 2020 position after its statute of limitations has expired

The agent's request. Please record uncertain tax position POS-2020-031 for Northgate Components, Inc., a 2020 US federal position supported by a regulation. Technical merit is 70 percent and the gross tax benefit is $500,000. Settlement outcomes: $500,000 at a cumulative 30 percent, $300,000 at 55 percent and $100,000 at 100 percent. Please record the position.

Verdict: Blocked.

  • Interest accrual missing. Interest has not been accrued on the $200,000 of unrecognized benefit.
  • Lapsed position not reversed. The statute of limitations for tax year 2020 has expired, but the reserve has not been reversed.

2. Lapsed statute of limitations: The agent resubmits after fixing the errors

The agent's request. Please record uncertain tax position POS-2020-031 for Northgate Components, Inc., a 2020 US federal position supported by a regulation. Technical merit is 70 percent and the gross tax benefit is $500,000. Settlement outcomes: $500,000 at a cumulative 30 percent, $300,000 at 55 percent and $100,000 at 100 percent. Please record the position.

Verdict: Blocked.

  • Lapsed position not reversed. The statute of limitations for tax year 2020 has expired, but the reserve has not been reversed.

3. Lapsed statute of limitations: The agent resubmits after fixing the errors

The agent's request. Please record uncertain tax position POS-2020-031 for Northgate Components, Inc., a 2020 US federal position supported by a regulation. Technical merit is 70 percent and the gross tax benefit is $500,000. Settlement outcomes: $500,000 at a cumulative 30 percent, $300,000 at 55 percent and $100,000 at 100 percent. Please record the position.

Verdict: Allowed.

  • Position summary. The position is recognized. Largest amount at least 50 percent likely: $300,000; reserve for the unrecognized portion: $200,000.

4. Effective settlement: A settled 2025 position with a settlement above the measured benefit

The agent's request. Please record uncertain tax position POS-2025-044 for Northgate Components, Inc., a 2025 US federal position supported by a regulation. Technical merit is 70 percent and the gross tax benefit is $500,000. Settlement outcomes: $500,000 at a cumulative 30 percent, $300,000 at 55 percent and $100,000 at 100 percent. The position has been effectively settled with the tax authority for $450,000. Please record the position.

Verdict: Blocked.

  • Settlement does not match the measured benefit. The effective settlement is $450,000; the measured benefit is $300,000.

5. Effective settlement: The agent resubmits after fixing the errors

The agent's request. Please record uncertain tax position POS-2025-044 for Northgate Components, Inc., a 2025 US federal position supported by a regulation. Technical merit is 70 percent and the gross tax benefit is $500,000. Settlement outcomes: $500,000 at a cumulative 30 percent, $300,000 at 55 percent and $100,000 at 100 percent. The position has been effectively settled with the tax authority for $300,000. Please record the position.

Verdict: Allowed.

  • Position summary. The position is recognized. Largest amount at least 50 percent likely: $300,000; reserve for the unrecognized portion: $200,000.

6. Weak authority: A 2025 position that rests on no direct authority

The agent's request. Please record uncertain tax position POS-2025-052 for Northgate Components, Inc., a 2025 US federal position with no direct authority. Technical merit is 60 percent and the gross tax benefit is $500,000. Settlement outcomes: $500,000 at a cumulative 30 percent, $300,000 at 55 percent and $100,000 at 100 percent. Please record the position.

Verdict: Needs approval.

  • Weak authority needs partner review. The position is recognized on no direct authority, so a partner must review it before it is recorded.
  • Position summary. The position is recognized. Largest amount at least 50 percent likely: $300,000; reserve for the unrecognized portion: $200,000.

7. Weak authority: The agent resubmits after fixing the errors

The agent's request. Please record uncertain tax position POS-2025-052 for Northgate Components, Inc., a 2025 US federal position supported by a regulation. Technical merit is 60 percent and the gross tax benefit is $500,000. Settlement outcomes: $500,000 at a cumulative 30 percent, $300,000 at 55 percent and $100,000 at 100 percent. Please record the position.

Verdict: Allowed.

  • Position summary. The position is recognized. Largest amount at least 50 percent likely: $300,000; reserve for the unrecognized portion: $200,000.

Narration transcript

Click to expand

The text of the narrated walkthrough, step by step.

  1. Welcome to a demo of Predictika's Validator. An AI tax agent prepares an uncertain tax position entry under ASC 740 (formerly FIN 48): the authority behind the position, its technical merit, the settlement outcome table and any lapse or settlement. Before it is recorded, the agent sends it to Predictika, which checks it against a declarative model of the recognition and measurement rules and returns a verdict, every reason and the figures it computed.
  2. I will start with the first request by the agent.
  3. The agent says. Please record uncertain tax position POS-2020-031 for Northgate Components, Inc., a 2020 US federal position supported by a regulation. Technical merit is 70 percent and the gross tax benefit is $500,000. The rest of the request is on screen.
  4. Before it goes out, the agent calls the validator.
  5. Blocked, with two findings, all in one answer. Interest has not been accrued on the $200,000 of unrecognized benefit. The statute of limitations for tax year 2020 has expired, but the reserve has not been reversed.
  6. The validator returned two blocking findings in one answer. A check that reads one field at a time would pass most of this; it shows only when the choices in the position are checked against each other.
  7. What this shows: for interest and reversal tied to a calculated reserve, a rules-based tool does this reliably only when someone is vigilant.
  8. Now the agent resubmits.
  9. The corrected request. It changes the interest accrued to yes.
  10. It calls the validator again.
  11. Still blocked, with one finding, all in one answer. The statute of limitations for tax year 2020 has expired, but the reserve has not been reversed.
  12. One finding cleared; still blocked on lapsed position not reversed.
  13. The agent resubmits once more.
  14. The revised request. It changes the reserve reversed to yes.
  15. It calls the validator once more.
  16. This time the validator clears it.
  17. The earlier finding is cleared.
  18. Two problems caught in one call, fixed in two passes, and a clean position. That is the value of Predictika's validator: catching mistakes before the position is taken.