Gains:
- Ability to combine tax and accounting tasks in a single, secure AI-supported workflow, from research to declaration
- Ability to audit each critical deliverable before release by embedding validation gates, quality checklists, and expert/professional approval points into the workflow
- Ability to turn the principles of confidentiality, current legislation, source integrity and signature responsibility into a permanent professional habit
Module Exam
1. When answering a tax question, artificial intelligence showed a specific Income Tax Law article and a tax ruling number. What is the best action to take before using this output in professional work?
- A) Opening the article and ruling number from the official source and confirming its content and current validity ✔
- B) Writing directly to the opinion as the model answers confidently and in detail
- C) Seeing that the number seems reasonable and passing it on without opening the source
- D) Telling the customer 'this is what the artificial intelligence says' and leaving the responsibility to the model
Explanation: Large language models can fake item and feature numbers as if they were real (hallucination). Tax legislation also changes constantly. Therefore, every article, notification and ruling shown cannot be used without opening it verbatim from the official source (Official Gazette, GIB, legislation portal) and confirming its content and current validity.
2. A financial advisor wants to analyze the payroll table containing the name, ID and salary information of a customer with 3,500 employees. Which is the most correct approach?
- A) Upload the data to the first free web tool it finds and request a summary
- B) Using corporate approved vehicles or anonymizing by masking identifiers such as name and TR ID ✔
- C) Just change the salaries a little and leave the names and IDs
- D) Sharing data as is and worrying about privacy
Explanation: Taxpayer data is protected by the confidentiality obligation of both the KVKK and the Tax Procedure Law. Uploading personal data to an uncontrolled, free tool is a violation. A corporate, approved tool that guarantees data retention should be used, or data should be anonymized by replacing identifiers such as name and TR ID with a code/mask.
3. What is the most critical verification step when artificial intelligence calculates a VAT base and tax payable?
- A) Trusting the model's fluent explanation and confident language and accepting the result
- B) Just looking at the result looking like a reasonable number
- C) Recalculating each base, rate and amount independently, manually or in a spreadsheet ✔
- D) Transferring the account directly to the declaration without checking it at all
Explanation: Large language models are powerful at generating text, but can silently make mistakes in arithmetic and confidently present the wrong result. Even a small mistake in tax calculation can lead to tax loss and penalties. Therefore, each base, rate and amount must be recalculated independently, either manually or with a table.
4. Why is 'currentness' a separate verification heading when using artificial intelligence in tax legislation research?
- A) Currentness is unimportant because the legislation has never changed.
- B) There is no need for control as the model always knows the most current legislation
- C) Just having the correct name of the law is sufficient for up-to-dateness.
- D) Legislation changes frequently and the effective date of each provision should be checked separately as the model's information may be outdated ✔
Description: Tax legislation changes very frequently: rates, exemption limits, notifications and practices can be updated even during the year. The model's training data is cut off at a specific date and can present a repealed provision as if it were still valid. Therefore, the effective date and amendment date should be checked separately for each provision.
5. What is the relationship of responsibility in preparing a declaration with artificial intelligence support?
- A) Artificial intelligence output can be directly converted into a statement, the responsibility lies with the model
- B) The responsibility lies with the signing professional and the taxpayer; AI is just a drafting tool that needs to be verified ✔
- C) When the declaration is prepared with artificial intelligence, there will be no errors, control is unnecessary.
- D) Responsibility belongs to the software provider
Explanation: The legal responsibility for the declaration lies with the professional and taxpayer who signed it. AI is just a tool for draft control, consistency, and forgotten pen reminders; The output cannot be converted directly to a declaration. Final figures must be confirmed by the official system and legislation and approved by a member of the profession.
6. Why is it misleading to interpret a ratio (e.g. current ratio) in isolation in financial statement analysis?
- A) There is no need to compare since the rates are always the same.
- B) A ratio is meaningless without industry benchmarks and periodic comparisons; Comments without context are misleading ✔
- C) A single ratio always reflects the entire financial position of the firm
- D) Only artificial intelligence should make the rate interpretation, humans should not interfere
Explanation: A ratio has meaning only with a benchmark: the industry average, previous periods, and the company's specific conditions. AI can label a single ratio as 'good' or 'bad'; but this interpretation is misleading without a sector and period comparison. Interpretation should always be made with context.
7. What is the most critical risk when drafting an objection petition to be submitted to the tax office with artificial intelligence and how is it managed?
- A) There is no risk; The model always produces legal texts correctly
- B) The only risk is the wording of the petition; no content control required
- C) The model may give a fitting item/period; Therefore, each basis should be confirmed from the official source, and the periods should not be left to a single basis ✔
- D) It is sufficient to ask the model for the times, additional control is unnecessary
Explanation: Artificial intelligence can fabricate a clause, notification or period that does not exist in the petition; Additionally, litigation and objection deadlines are detrimental. Therefore, every legal basis in the draft should be confirmed from the official source, and artificial intelligence should never be the sole basis for time-bound rights.
8. Which of the following is the appropriate role of artificial intelligence in tax audit preparation?
- A) Taking the place of the taxpayer in the examination and carrying out the official defense alone
- B) Being a preparation tool that compiles documents, maps consensus and criticism points, and whose output is verified ✔
- C) Establishing an automatic channel to respond directly to the auditor
- D) Leaving the defense entirely to the model without collecting any documents
Description: Artificial intelligence; It is a powerful preparation tool for compiling documents, reconciliations of amounts, mapping possible points of criticism, and drafting a defense argument. However, the final defense and declaration are the responsibility of the professional and the taxpayer; Every argument must be supported by legislation and documentary evidence.
9. What is the role of artificial intelligence output in a complex issue such as transfer pricing and subject to CPA certification?
- A) It replaces the CPA certification, the report can be prepared directly to the model.
- B) Artificial intelligence cannot be used at all in this regard because it is complex
- C) If there is an error, the responsibility passes to the model provider.
- D) Produces skeleton and draft, but approval, approval and final responsibility remain with the competent expert ✔
Description: On such issues, artificial intelligence can produce report frameworks, method comparisons and document drafts; But the CPA report does not replace certification and official approvals. The output should be confirmed by current legislation, circulars and established practice, and the final responsibility should remain with the competent expert.
10. When writing a good tax/accounting prompt to artificial intelligence, which element increases the output quality the most?
- A) Keeping the prompt as short and general as possible, e.g. just 'calculate tax'
- B) Be clear about the role, task, resource given, period/year, format and boundaries (prohibition of fabrication, tag to be verified) ✔
- C) Asking the model to fill in the missing information with its own prediction.
- D) Requesting a general account without specifying which year it is
Explanation: Vague demands lead the model to guess and exaggerate. Being clear about the role, task, given source (legislation text, table), period/year, format and boundaries ('rely only on the source I gave, make up the article, mark if you are not sure should be verified') makes the output both accurate and safe.
11. What does 'validation gate' (checkpoint) mean in an end-to-end AI-supported tax workflow?
- A) A setting that turns off artificial intelligence's internet access
- B) A button that allows declarations to be sent automatically
- C) The point at which a critical output is checked and approved by the professional against source, account and legislation before moving on to the next step ✔
- D) Backup feature of accounting software
Description: The verification gate is the point where an AI output is checked and approved by the professional against the source, account and legislation before moving on to the next step (especially the declaration, official correspondence or customer). Critical outputs such as tax base, legislative reference, petition and declaration are not published without passing through these gates; The process becomes traceable.
12. Why is it risky to directly rely on a tax rate or exemption amount in the AI's memory?
- A) Rates and exemption amounts change frequently; The remembered value of the model may be outdated and should be verified from official source ✔
- B) Since the proportions never change, what the model gives is always true
- C) The model automatically updates the rates every day
- D) Instead of verifying the ratio, it is sufficient to use an approximate value
Explanation: Tax rates, exemption and deduction limits change almost every year (sometimes within the year). The model's information is frozen in the training date and may return an old rate as if it were current. Therefore, periodic parameters should always be verified from the official source for that year, and the value remembered by the model should not be trusted.
13. What is the most common pitfall when creating a compliance checklist with AI?
- A) Use the list as an initial draft and link it to legislation
- B) Matching each item with the actual situation of the firm
- C) Using the list as it is, with general and outdated items, without linking it to the legislation and the company's reality ✔
- D) Adding a human approval gate to critical steps
Explanation: Artificial intelligence can produce a list full of general, inappropriate or outdated items. The value is in tying each item to the relevant legislation and the real situation of the company and adding a human approval gate to each critical step. Automation does not eliminate responsibility.
14. What is the consequence when a tax error (for example, incorrect tax base declaration) is made based on artificial intelligence output and how does this affect the verification discipline?
- A) No consequences arise; responsibility is in the model
- B) Only the model gives a warning, additional control is unnecessary
- C) Only a small warning comes, no penalty or interest
- D) Tax loss, penalty, interest and professional liability may arise; so every critical output must be verified as source, account and logic ✔
Explanation: Misrepresentation may result in tax loss, tax loss penalties, late interest and joint liability for the professional. 'Artificial intelligence calculated it that way' is not a defence. These heavy consequences precisely make it necessary for every critical output to be verified in terms of source, calculation and logic; unverified output is an unsigned draft.