Evidence informs. Humans decide. The system preserves the reasoning.
One decision, traced from its source documents to a record anyone can reproduce: the assumptions it rests on, the arithmetic, the case against it, the places the documents disagree and what still needs a person to decide.
Synthetic demonstration The borrower, lender, documents and figures are fictional. No bank, fund or institution has reviewed, validated or used this. The calculations are deterministic rules, not AI, and not a credit model: no probability of default is estimated and nothing here is credit advice or an approval.
AbrahamMoses is a scholarly research record. Sagolik Institutional is a proposed commercial application of the same method (sources, assumptions, dependencies, review) to institutional decisions. It has not been validated by any bank or institution. More on the difference.
1 · Decision
Should the committee approve the proposed facility under the specified conditions?
Exemplar Industrial Holdings (fictional) asks for a $120M senior secured facility to refinance existing bank debt and fund a plant expansion. Case DT-PC-001 v1.0.0.
Facility structure
Tranche
Amount
Terms
Term loan
$90M
Five years, amortising 5% a year
Revolving credit facility
$30M
$10M drawn at close
Decision configuration
Configuration dr-config-1.0.0. Saved into every Decision Passport with the assumption values below, and covered by its hash, so the record can be reproduced. In a private workspace these would be set per decision; here they are fixed for the synthetic case.
Objective
Decide whether the synthetic $120M facility can go to a (simulated) credit committee, and on which conditions.
Evidence standard
Every figure used in a calculation must trace to a named source document. Unverified documents may inform the analysis but cannot by themselves satisfy a condition; the verification condition stays unmet until they are verified by a human.
Approval policy
propose / review / reject / approve. The system may only propose. Review, rejection and approval are human actions by authorized people. In this public demonstration no human action is recorded and nothing is approved.
Escalation rules
A condition with missing evidence (EVIDENCE_MISSING) is escalated to a human before any recommendation.
A confirmed contradiction between sources is escalated until a human resolves which source governs.
An unresolved interpretation of a contract term is escalated to legal review.
A different reporting basis is shown side by side and is not escalated as a conflict.
Monitoring triggers
A new or revised source document for this case (planned; no monitoring runs in the demonstration).
Top customer renewal outcome in Q1 2027 (DOC-07).
Collateral appraisal received (DOC-08).
Covenant headroom below the configured threshold in a new compliance certificate.
Assumption states: fixed, adjustable and disputed
State
Assumption
Value
Source
fixed
Revenue FY2025 Taken from the source document; not adjustable in this case.
$260M
DOC-01
fixed
Reported EBITDA margin FY2025 Taken from the source document; not adjustable in this case.
16%
DOC-01
fixed
Cash at close Taken from the source document; not adjustable in this case.
$15M
DOC-03
fixed
Undrawn RCF Taken from the source document; not adjustable in this case.
$20M
DOC-03
fixed
Minimum interest cover Taken from the source document; not adjustable in this case.
3x
DOC-05
fixed
Customer concentration policy limit Taken from the source document; not adjustable in this case.
25%
DOC-04
fixed
Stress EBITDA decline Taken from the source document; not adjustable in this case.
20%
DOC-04
adjustable
Revenue growth FY2026 Source value.
12%
DOC-02
adjustable
EBITDA margin Source value.
17.5%
DOC-02
adjustable
Total debt at close Source value.
$124M
DOC-03
adjustable
Annual interest expense Source value.
$10.5M
DOC-03
adjustable
Leverage covenant ceiling Source value.
3.5x
DOC-05
adjustable
Minimum liquidity requirement Source value.
$10M
DOC-05
adjustable
Top customer share of revenue Source value.
22%
DOC-07
disputed
P1: Management’s 12% revenue growth for FY2026 is achievable. If rejected: Revenue growth falls back to FY2025 actual growth of 4.0% (DOC-01).
accepted
DOC-02
disputed
P2: Management’s 17.5% EBITDA margin for FY2026 is achievable. If rejected: EBITDA margin falls back to the FY2025 reported margin of 16.0% (DOC-01).
accepted
DOC-02
disputed
P3: The top customer renews on current terms in Q1 2027. If rejected: Revenue loses half of the top customer’s share (dual-sourcing, DOC-07).
accepted
DOC-02, DOC-07
Default workspace settings (planned for private workspaces; read-only)
Planned defaults for a private Decision Twin workspace. No private workspace exists yet; nothing here can be changed and none of it is in effect on this public site.
Each document is structured data with a stable ID. The SHA-256 hash of its canonical JSON lets a passport show which exact version it relied on. “Verified” is the synthetic data room’s own flag; you can change it under Assumptions.
Reconcile total debt between the memo and the debt schedule.
—
—
Q2
Which EBITDA definition will the covenant use?
—
—
Q3
Provide the PP&E appraisal.
—
—
Q4
What happens if the top customer dual-sources?
—
—
3 · Assumptions
Change an input or reject a premise
Recalculating reloads the page with your inputs in the address, so every view is a link that reproduces itself. Only what depends on a changed input moves; everything else stays as it was.
Total debt reconciled between memo and debt schedule
Needs human judgment
memo $118.0M vs schedule $124.0M; calculations use $124M
DOC-03, DOC-04
Covenant EBITDA definition agreed
Needs human judgment
draft agreement has no add-back cap; term sheet caps add-backs at 10%
DOC-05, DOC-06
Collateral value supported by the cited appraisal
Evidence missing
appraisal A-2026-03 is not in the data room
DOC-08
Warnings
Active: Forecast growth well above history and sector outlook: 12.0% growth against 4.0% in FY2025 and a 4.5% sector outlook
Clear: Thin covenant headroom (under 25%)
Clear: Stress case coverage below 3x
Active: Forecast relies on the top customer renewing: 22% of revenue renews in Q1 2027; the customer has signalled dual-sourcing
Active: Calculations rely on unverified documents: DOC-03, DOC-05 not verified
6 · Adversarial Chamber
The case for, the case against, and what the evidence can bear
Deterministic rules, not AI. Each side is a fixed rule set applied to the same conditions, premises and documents shown above. Reject a premise under Assumptions and every side is recomputed. No argument here is a human review.
Supporting case
Rules: A condition that is met supports the facility. An accepted premise is cited with its source.
Debt / EBITDA within the leverage covenant: 2.43x ≤ 3.5x.DOC-05 · k.leverage · m.leverage · in.covMax
Interest coverage at least 3.00x: 4.85x ≥ 3x.DOC-05 · k.coverage · m.coverage
Forecast growth well above history and sector outlook: 12.0% growth against 4.0% in FY2025 and a 4.5% sector outlook.DOC-01 · DOC-02 · DOC-09 · w.forecastVsHistory · e.growth
Forecast relies on the top customer renewing: 22% of revenue renews in Q1 2027; the customer has signalled dual-sourcing.DOC-02 · DOC-07 · w.renewal · prem.P3 · in.concentration
Calculations rely on unverified documents: DOC-03, DOC-05 not verified.DOC-01 · DOC-03 · DOC-05 · w.unverified · in.verified
Evidence auditor
Rules: Every unverified key document is flagged with what relies on it. Every conflict, missing item and unresolved interpretation is listed. Differences of reporting basis are named and explicitly not counted as conflicts.
DOC-03 (Pro forma debt schedule at close) is unverified; 7 calculation(s) or condition(s) cite it.DOC-03 · m.leverage · m.coverage · m.liquidity · m.stressLeverage
Confirmed conflict X-01: Total debt at close. Reconcile the two figures before the committee relies on either. The calculations default to the higher, itemised figure.DOC-03 · DOC-04 · in.debt · m.leverage · m.stressLeverage · k.leverage
Potential conflict X-02: Top customer volumes. Ask management for evidence of the customer’s FY2026 orders; test the case with premise P3 rejected.DOC-02 · DOC-07 · prem.P3 · m.revenue · w.renewal
X-03 (EBITDA: $41.6M vs $46.8M) is a difference of reporting basis, not a conflict.DOC-01 · DOC-04
Missing evidence X-04: PP&E appraisal not provided. Obtain the appraisal (committee question Q3).DOC-08 · k.collateral
Unresolved interpretation X-05: Which EBITDA the leverage covenant uses. Counsel and the deal team agree the definition; the draft is conformed to the term sheet or the difference is approved.DOC-05 · DOC-06 · k.ebitdaDefinition
7 · Contradiction Map
Where the documents disagree, and where they only seem to
Five classes. Only confirmed and potential conflicts are conflicts. A difference of period, scope, definition or version is a different reporting basis and is not called a contradiction.
X-01 Total debt at close Confirmed conflict
Side-by-side comparison for X-01
Document
Field
Value
Period
Basis
DOC-03 Pro forma debt schedule at close
Total debt
$124.0M
At close
Total debt incl. leases and seller note
DOC-04 Credit memo (draft for committee)
Total debt
$118.0M
At close
Total debt incl. leases and seller note
Same measure, same date (at close), same stated basis (including leases and seller note), different amounts ($124.0M vs $118.0M). The $6.0M gap is unexplained.
Needs: Reconcile the two figures before the committee relies on either. The calculations default to the higher, itemised figure. Bears on in.debt, m.leverage, m.stressLeverage, k.leverage, k.stress, k.debtReconciled.
X-02 Top customer volumes Potential conflict
Side-by-side comparison for X-02
Document
Field
Value
Period
Basis
DOC-02 Management forecast FY2026
Top customer volume growth
15%
FY2026
Management case
DOC-07 Customer concentration report
Stated intent
Dual-sourcing under consideration
—
—
The forecast assumes 15% volume growth from the top customer; the concentration report says that customer is considering dual-sourcing at its Q1 2027 renewal. Both can be true for FY2026, so this is a potential, not a confirmed, conflict.
Needs: Ask management for evidence of the customer’s FY2026 orders; test the case with premise P3 rejected. Bears on prem.P3, m.revenue, w.renewal.
X-03 EBITDA: $41.6M vs $46.8M Different reporting basis (not a conflict)
Side-by-side comparison for X-03
Document
Field
Value
Period
Basis
DOC-01 Audited financial statements FY2025
EBITDA FY2025
$41.6M
FY2025
Reported (no add-backs)
DOC-04 Credit memo (draft for committee)
Adjusted EBITDA LTM
$46.8M
LTM to 30 Jun 2026
Adjusted (incl. $3.2M add-backs)
Different period (FY2025 vs LTM to 30 June 2026) and different definition (reported vs adjusted with $3.2M add-backs). The figures measure different things, so the difference is not a contradiction. The independent note (DOC-09) agrees with the reported $41.6M.
Needs: None required for consistency. Which definition the covenant uses is a separate question (X-05).
X-04 PP&E appraisal not provided Missing evidence
Side-by-side comparison for X-04
Document
Field
Value
Period
Basis
DOC-08 Collateral summary
PP&E orderly liquidation value
$52.0M
—
Per appraisal A-2026-03 (not provided)
No document in the data room supports this value.
The collateral summary cites appraisal A-2026-03 for the $52.0M orderly liquidation value, but the appraisal is not in the data room.
Needs: Obtain the appraisal (committee question Q3). Bears on k.collateral.
X-05 Which EBITDA the leverage covenant uses Unresolved interpretation
Side-by-side comparison for X-05
Document
Field
Value
Period
Basis
DOC-05 Draft loan agreement (financial covenants)
Cap on non-recurring add-backs
None
—
Clause 22.1 definition
DOC-06 Covenant definitions schedule (term sheet)
Cap on non-recurring add-backs
10% of EBITDA
—
Term sheet
Both documents set the same 3.50x ceiling, but the draft agreement adds back non-recurring charges without a cap while the term sheet caps them at 10% of EBITDA. Whether restructuring costs qualify is a drafting and legal question, not an arithmetic one.
Needs: Counsel and the deal team agree the definition; the draft is conformed to the term sheet or the difference is approved. Bears on k.ebitdaDefinition.
8 · Reasoning graph
What depends on what
Every derived node and the nodes it reads. A change propagates only along these edges; the test suite checks that nothing outside them moves.
This demonstration makes no decision. In a real deployment a named committee would record its decision, its reasons and any conditions, and that record would be added to the passport. Here the reviewer is none (simulated decision).
Needs human judgment on these inputs
Total debt reconciled between memo and debt schedule: memo $118.0M vs schedule $124.0M; calculations use $124M
Covenant EBITDA definition agreed: draft agreement has no add-back cap; term sheet caps add-backs at 10%
Collateral value supported by the cited appraisal: appraisal A-2026-03 is not in the data room
10 · Decision Passport
A record anyone can reproduce
The passport records the sources and their hashes, every input and premise, the calculations, conditions, conflicts, open questions and the audit trace, with a SHA-256 integrity hash. Its address encodes all inputs, so it reloads identically without a database. Ruleset dr-rules-1.0.0; no AI model.
Read-only API for this demonstration: GET /api/institutional/decision-room/case, /scenario, /contradictions and /passport/verify, described in the OpenAPI document (tag institutional-demo).
Defaults: Revenue growth FY2026 12% · EBITDA margin 17.5% · Total debt at close 124$M · Annual interest expense 10.5$M · Leverage covenant ceiling 3.5x · Minimum liquidity requirement 10$M · Top customer share of revenue 22%.