FICTIONAL CPA READINESS EXERCISE Companion reader asset for the November 2026 evidence-readiness series Prepared November 6, 2026. NOT RUN in TaxFlo, SafeVault, or a customer workflow. No taxpayer records, actual account numbers, tax forms, taxpayer identifiers or live access are required. All names and amounts below are invented for this exercise. Purpose Test whether a review process notices a source correction, an incorrect engagement association and missing expected evidence, including items outside an AI exceptions queue. This exercise does not decide tax treatment, deductibility, eligibility or a return adjustment. A practitioner must approve the answer key before use. Packet to prepare Engagement J: Juniper Repairs LLC, calendar year 2025, an invented business. Engagement P: Juniper Property LLC, calendar year 2025, a separate invented business. Records have human-readable IDs and explicit entity, period and revision labels. Do not use real bank logos, taxpayer details or production accounts. Store source versions separately rather than overwriting them. Case A — current ordinary record J-01-v1: Juniper Repairs LLC, January 2025, invented supplier statement, total $1,200. No later revision is supplied. A worksheet references the same record ID and period. Expected evidence-readiness observation: the supplied entity, period and source association agree. This does not prove the underlying expense is allowable, that the source is authentic, or that the full engagement is complete. No invented discrepancy should be required to demonstrate activity. Case B — later correction J-02-v1: Juniper Repairs LLC, February 2025, invented supplier statement, $2,400. The initial review note references v1. J-02-v2: same entity/period/source family, explicitly marked correction replacing v1, $2,100, received after the initial review note. Expected observation: identify the corrected version, retain the earlier version and make the earlier review's dependency visible. Ask the responsible practitioner whether the earlier disposition needs review. Do not silently replace the review note or prescribe a journal entry. An extracted $300 difference alone is insufficient closure. Case C — plausible wrong entity P-03-v1: Juniper Property LLC, March 2025, invented statement, $1,700. Place a reference to it in Engagement J's packet, with a similar file name and the same month. Expected observation: the source identifies a different entity. Do not treat a matching amount/month or similar name as authorization to use it in J. Leave the association unresolved for practitioner review; do not automatically move a live record or expose one client to another. Case D — absent expected evidence J-04-checklist: an invented current-year checklist says a closing balance needs supporting statement J-04. Supply a follow-up note referring to the statement but do not supply J-04. Expected observation: the note is not the missing statement. Keep the evidence question open, name its owner and identify what is still requested. Do not invent a zero balance, document or completed follow-up. Missing evidence does not itself determine a tax conclusion. Case E — prior-period context J-05-v1: Juniper Repairs LLC, December 2024, invented statement, $900. Engagement J's task asks for 2025 evidence. The label makes the older period explicit. Expected observation: distinguish prior-period context from current-year support. Authorization to use prior-year information must be established in any real workflow. Do not assume the same account, balance or transaction continues in 2025. A prior-year reference can prompt a current-year question, not close it. Independent comparison protocol Before seeing any AI flags, the practitioner chooses A, C and E for a separate read and writes observations against this answer key. Then review all five cases through the firm's existing process and, if separately authorized, the proposed assisted process. The independent selection here is deliberate and small; it is not a statistically representative sample or a production miss-rate estimate. For each case record: input IDs/versions; expected evidence question; system output or no flag; independent observation; unsupported flag; missed question; unsupported closure; appropriate unresolved disposition; owner; inspected evidence; preparer minutes; reviewer minutes; follow-up minutes; rework minutes. Actual observations and timings are BLANK until execution. Do not prefill PASS, accuracy or time saved. Version-change comparison Preserve the packet, practitioner-approved key, selected independent cases, task instructions and observed results. Record product/release identifier, date, permitted context and reviewer. Repeat after a relevant change only in an authorized test setting. If the earlier version is unavailable, identify the missing controlled baseline. Do not claim universal rollback or that a wording change proves a regression. Record accepted differences, unresolved differences and the person authorizing any next step. Commercial next step Use a 20-minute fictional workflow review to learn whether any case resembles a real handoff pain and agree a useful measure. Only then consider a controlled 10–25 engagement pilot, with the practice's authorization and appropriate data controls. This asset is not evidence that a review, pilot or product test occurred. Editorial continuity and sources Published continuity: https://taxflo.us/blog/cpa-ai-document-client-period-association-test Primary resources rechecked November 6: https://www.cpa.com/sites/cpa/files/media/resources/whitepapers/ai-solution-due-diligence-guide-for-accounting-firms-cpacom.pdf https://airc.nist.gov/airmf-resources/airmf/5-sec-core/ These resources support asking about vendor validation, changes and ongoing evaluation. They do not prescribe this five-case exercise, certify AmitaSoft or establish measured results. The exercise and answer key above are AmitaSoft's proposed reader asset.