Skip to main content

Changelog

2026-10-16: GAAPx Score, new stage names, eight areas​

  • The score is now called the GAAPx Score. It is still a whole number from 0 to 100, and it is self-assessed. The product is now called GAAPx. Earlier entries below keep the names used at the time.
  • Stage 3 is now Agent-assisted (was AI-assisted) and Stage 4 is now Agentic (was AI-native). The stages page is titled "Five stages, from Manual to Agentic". Bands and scoring are unchanged. Agentic still means a named person owns each control. Nobody requires you to reach Stage 4. Old anchors on the stages page still work.
  • Beyond Stage 4 is shorter. It says the fully agentic enterprise is not here yet. The line about agents from other companies is removed.
  • Pay Ops is now Spend (accounts payable). Spend excludes payroll.
  • Payroll is added as a new area, called Payroll and benefits in plain terms, placed before The Close. The area list is now eight. Its workflows are Pay run, Starters and changes, Benefits and deductions, and Journal and reconciliation. Payroll use cases show as coming soon.
  • There are now 32 scored questions, four per area. The GAAPx Score is the mean of the eight area scores x 10, so it still runs from 0 to 100. Each area carries 1/8 of the score instead of 1/7. The worked examples on how the score works are redone for eight areas.
  • Results saved under the seven-area list are not comparable with eight-area results. Compare a result only with others from the same version.
  • Addresses of docs pages are unchanged. The "seven-areas" page keeps its address and is now titled "The eight areas".

2026-10-14: area labels, two scores​

  • Each area now shows its branded name and its plain term together: Pay Ops · Accounts payable, Get Paid · Order to cash, The Close · Month-end close, Plan vs Actual · Budgeting and FP&A, Cash Ops · Treasury and cash, Controls · Internal controls, Reporting · Financial reporting.
  • Control Room is renamed Controls. Area ids and addresses are unchanged.
  • The Combined score is removed. You get two scores, the Benchmark score and the Readiness score. Readiness is separate and never changes the Benchmark score. The start gate at 60 is unchanged: both scores are read separately against it.

2026-10-11: stages are numbered 0 to 4​

  • The five stages now carry numbers wherever they appear: Stage 0, Manual. Stage 1, Digitized. Stage 2, Automating. Stage 3, AI-assisted. Stage 4, AI-native. The name always appears with the number. The bands and the scoring are unchanged. The anchors on the stages page (for example "#manual") still work.

2026-10-11: top stage renamed, score ring​

  • The top of the five stages is now called AI-native (was AI-operated). The five stages are Manual, Digitized, Automating, AI-assisted and AI-native. The band (80 to 100) and the description are unchanged. The anchor and the page link on the stages page now use "ai-native".
  • Score rings now show the number large with a small, muted "/100" beneath it. The score is unchanged.

2026-10-11: seven areas, clearer definitions, governance structure​

  • The area list is now seven. Reporting is added as area 7. It covers internal reporting (management and board slides and packs) and external reporting (financial statements, footnote disclosures, statutory filings, and lender and audit-facing reports). It has three workflows: Management reporting, Statutory, regulatory and external filings, and Board and investor reporting.
  • The Close is narrowed to the period-end close: entries and accruals, reconciliations and sign-off. Producing reports from the closed books moves to Reporting.
  • Areas are renamed: Pay Ops (was Procure-to-pay), Get Paid (Order-to-cash), The Close (Record-to-report), Plan vs Actual (Plan-to-perform), Cash Ops (Treasury and capital), Control Room (Controls, risk and assurance). The technical names now appear only in "What it covers".
  • Boundaries: the cash flow forecast stays in Cash Ops and variance analysis stays in Plan vs Actual. Reporting publishes both.
  • The benchmark score is now the mean of seven area scores x 10, a whole number out of 100, based on 7 of 7 areas. Bands, ratios and weights are unchanged.
  • Saved six-area results are not comparable with seven-area results. Each area now carries 1/7 of the benchmark instead of 1/6, so the same answers can give a different number. Compare a result only with others from the same version.
  • Definitions clarified. Steps count the planned steps and do not count rework. Loops count every trip back to an earlier step to fix something, and the loop count is still reported so people can see it fall. Break time is counted in working days, so weekends and holidays are not counted. "Not sure" counts as the lowest band.
  • The weights are a starting point and will change as practitioners weigh in.
  • The product is now called GAAPx Benchmark.
  • Governance structure added, informed by the NIST AI Risk Management Framework. Four functions (Govern, Map, Measure, Manage), each at two levels (Intent and Tested), eight draft questions. The four statuses are unchanged. Governance is asked once for the whole finance function and never changes the benchmark score. There is no claim of certification, alignment, compliance or endorsement. See NIST AI RMF mapping.

Rules version efficiency-v1 (proposed)​

  • New measure: workflow efficiency. Each workflow is scored on steps, loops, owners and break time. Break time counts double.
  • The benchmark score comes from one set of answers per area. Four scored questions (steps, loops, owners, break time), each a band from 1 to 5 that maps to a ratio (5 = 1.0 down to 1 = 0.2). Area efficiency = (steps + loops + owners + 2 x break) / 5. Area score = efficiency x 10. Benchmark score = average of the six area scores x 10, out of 100.
  • The benchmark score is always based on 6 of 6 areas.
  • A skipped question or "Not sure" is allowed and counts as the lowest level (band 1, 20 percent). Weights stay the same and every area is scored. The report says how many answers were left blank.
  • The workflow-level formula stays as a deeper view for up to three workflows the reader picks. The standard ideal is 6 steps (was 5), and can differ per flow later. Workflow diagrams are drawn from the steps the reader documents.
  • Six areas replace the ten. Tax, and Payroll and employee expenses, are left out for now.
  • The five dimensions and the workflow score are retired.
  • Break-time ideals are assumptions until researched with practitioners. This is proposed and may change.
  • The journey scale and stages are unchanged.

Version 0.1​

What exists now:

  • The five dimensions with proposed weights.
  • The automation and AI journey scale, 0 to 10, with roll-up rules (journey-v1).
  • The ten areas: nine pillars plus Governance.
  • Governance statuses and a note on the NIST AI RMF as a voluntary reference.
  • Three metric definitions: invoice cycle time, first-pass match rate and days to close.
  • Use-case Impact and Ease scoring.
  • Contributor information.

The method is proposed and will change. Each change will be listed here.