Send a workbook
Financial model review for Excel

Catch the cell that could change the decision.

Second Review checks finance workbooks before they reach a client, board, or lender. You get a prioritized PDF with the exact cell, the evidence, the likely impact, and the next step. Your model stays untouched.

Get a second reviewSee sample findings

Pilot review from $249 per workbook · 48-hour turnaround

Exact cell evidenceNo automatic editsDeterministic checks + AI review
WORKBOOK REVIEW
13-Week Cash Forecast.xlsx
Review ready
12reviewed findings
3high priority
100%exact locations
Cash Forecast
HighCash Forecast!N42

Hardcoded 25% breaks the row’s formula pattern. It feeds minimum-cash and covenant outputs and needs confirmation.

A second pair of eyes, with an evidence trail.
01Read-only
workbook handling
02Cell-level
findings and context
03Human
review stays in control
How it works

From workbook to defensible review in three steps.

No setup project and no new spreadsheet workflow. Send the file you already need to sign off, then use the report to review the risky cells first.

48hpilot turnaround
01Your input

Send one .xlsx workbook

A monthly pack, cash forecast, budget-vs-actual, covenant model, or consultancy model. Source documents or a prior version can support reconciliation checks.

02The review

Verification runs in two layers

Deterministic rules find formula, reconciliation, and consistency failures. AI reviews grounded evidence to remove noise, prioritize, and explain.

03Your output

Open a prioritized PDF report

Every issue includes its location, evidence, impact, confidence, and next step. You decide what to accept or dismiss; the workbook is never changed.

Audit coverage

Checks that go beyond “the formula runs.”

Excel can calculate a wrong model perfectly. Second Review looks for structural breaks, conflicting inputs, failed reconciliations, and evidence that does not support the output.

ƒx

Formula integrity

Where the model stops behaving like its surrounding structure.

  • Broken row and period patterns
  • Pasted hardcodes in formula regions
  • Numeric constants inside formulas
  • Broken references and masked errors
Σ

Reconciliation

Whether totals and linked statements agree with their components.

  • Totals and subtotals that fail
  • Cash and balance inconsistencies
  • Cross-sheet mismatches
  • Source document discrepancies
⇄

Model consistency

Values that conflict across tabs, periods, and conventions.

  • Conflicting assumptions
  • Mixed units, signs, and currencies
  • Date convention changes
  • Duplicate or missing records
◎

Decision impact

How a suspicious cell connects to outputs that matter.

  • Downstream dependent cells
  • Material output sensitivity
  • Unusually influential changes
  • Evidence needed to confirm
Sample findings

Each finding answers the reviewer’s next five questions.

What is wrong? Where is it? What proves it? What could it affect? What should I check next? The report keeps those answers together.

FINDING · FORMULA-PATTERN-001CONFIDENCE · 96%
High · needs confirmationCash Forecast!N42

A hardcoded rate breaks the forecast row’s formula pattern.

N42 contains 25%, while periods J–M and O–T reference Assumptions!B5. It may be intentional, but it changes the covenant forecast and needs confirmation.

Failed comparison

J42:M42 → Assumptions!B5
N42 → 25%

Downstream effect

Feeds Covenants!D18 and Summary!C7.

Why it matters

Forecast headroom falls from 1.34x to 1.21x.

Reviewer status

Open · awaiting model-owner confirmation.

Suggested action: confirm whether the 25% override is approved. If not, restore the driver reference and rerun the covenant output.
  1. Priority and confidence

    What to review first and how strong the evidence is.

  2. Exact location

    Sheet, cell, or range for immediate verification.

  3. Rule and evidence

    The failed comparison, pattern, or reconciliation.

  4. Downstream impact

    The cash, covenant, margin, or output affected.

  5. Recommended investigation

    A concrete step with the final decision left to you.

HighSummary!F18

Cash balance does not reconcile to the forecast schedule.

The summary is $184,200 above detailed ending cash for the same period.

Evidence: linked totals · Impact: board liquidity view
MediumAssumptions!D9

Revenue growth assumption disagrees across tabs.

The operating model uses 8.5%, while the board case still references 6.0% for FY27.

Evidence: cross-sheet contradiction · Action: confirm source of truth
Human reviewDebt!M31:M36

A formula change may be an intentional refinancing step.

Context is insufficient to call it an error, so the report keeps it open for confirmation.

Evidence: formula transition · Status: not silently removed
The category

A second reviewer, not another spreadsheet author.

Rigorous cell-level checks for recurring finance work, without rewriting your workbook or pretending to provide audit assurance.

AI Excel copilots Second Review Model audit firms
Role Builds and edits. Reviews and cites evidence. Provides formal assurance.
When While you build. Before the pack or forecast ships. Usually around a transaction.
Output A changed workbook. Prioritized evidence-linked PDF. Opinion letter after full review.
Control Tool may edit cells. You decide; the file stays untouched. Reviewer-led engagement.
Cost $20–200/mo From $249/workbook $25,000–50,000/engagement

Second Review assists a qualified reviewer. It does not provide audit assurance, a reliance letter, accounting advice, or investment advice.

Who it is for

Built for people who sign off on someone else’s spreadsheet.

A strong fit if you are

  • A fractional CFO reviewing monthly packs
  • A controller or FP&A lead reporting to a board or lender
  • A consultant shipping a model with your name on it
  • Accountable for a workbook you did not build

Not the right tool if you need

  • An AI to build or rewrite the model
  • Audit assurance or a reliance letter
  • Regulated model-risk validation
  • A button that changes cells automatically
✓Every claim links to a finding ID
✓Unsupported checks become visible limitations
✓Deterministic fallback remains usable
“The value is not another score. It is knowing exactly which cell to review first—and why.”
Launch pricing

Start with the workbook already on your deadline.

Every pilot review is machine-checked and human-verified before it reaches you.

Pilot review

$249 / workbook
  • Full two-layer review
  • Evidence-linked PDF report
  • 48-hour turnaround
  • 30-minute walkthrough
Send a workbook

Firm pilot

$950 / 3 workbooks
  • Three reviews in one month
  • Your firm’s house rules
  • Priority turnaround
  • Findings-quality debrief
Start a firm pilot

Monthly desk

from $500 / month
  • Recurring close-calendar reviews
  • Multi-client workbook queue
  • Finding history across periods
  • Volume pricing
Talk to us

Launch pricing may change. Pilot reviews remain human-verified while the product earns trust and builds its validation corpus.

Questions

The questions a careful reviewer asks.

What happens to my file?

Your workbook is never used to train models and is not changed. Before a production engagement, the exact processing location, AI provider, transmitted context, access controls, retention period, and deletion process should be confirmed in the engagement terms.

Is this an audit?

No. This is a second review that helps a qualified professional catch material errors. It does not provide formal assurance or a reliance letter.

How do you handle false positives?

Deterministic findings provide factual evidence. Grounded AI review can remove an issue only when supplied workbook context supports a false positive or intentional design. If context is insufficient, the finding remains open.

Why not paste it into ChatGPT or Copilot?

A general chat response is commentary, not a systematic verification pass. Second Review runs defined checks and gives every user-facing claim an exact workbook location.

What files are supported?

One .xlsx workbook today. VBA, external links, unsupported formulas, and unavailable capabilities are reported as limitations rather than silently skipped.

Does the tool fix the model?

No. It explains the issue and investigation step, but never writes into the workbook. You stay responsible for accepting, dismissing, or resolving each finding.

Your next review

The pack is going out either way. Send it with evidence behind it.

Get a second review of the cells, formulas, and reconciliations that could change the decision.

Get your workbook reviewed