IDALLOS
Book a scoping call

Independent decision assurance

One decision you can’t defend can undo a hundred you got right.

IDALLOS independently examines and assures consequential decisions — made by people, models, or agents — and produces the record showing what was known, who held authority, and which rules were in force at the time.

What we reconstruct, for every decision
  1. Evidence
  2. Authority
  3. Governing criteria
  4. Decision
  5. Assurance record
01 The moment

The challenge always arrives after the fact.

By the time a regulator, a plaintiff, or your own board asks how a decision was made, the evidence has scattered — across email, several systems, a contract nobody re-read, a policy that has since been rewritten, and the memory of someone who has since left. You spend weeks reconstructing under a deadline what should have been a matter of record. And by the time you do, someone else’s version of events is already the one on the table.

A log tells you an event happened. It does not tell you whether the decision behind it was defensible.

02 The part most teams miss

The absence of a record isn’t neutral. It’s the finding.

Where the law expects reasonable care — a fiduciary duty, an unfair-practices regime, a safety-and-soundness examination — you do not get credit for a good decision you cannot prove you made carefully. The exposure was never only the decision that went wrong. It is every consequential decision you cannot reconstruct. When the contemporaneous record doesn’t exist, its absence is what gets weighed against you.

You will not be judged on the decision alone. You will be judged on what you can show you knew when you made it.
03 Why now

Your systems now decide faster than anyone can reconstruct.

When an automated or AI system denies a claim, blocks a payment, or files an entry, it doesn’t record why it was allowed to — what authority it held, which policy version governed, what it should have escalated instead. Each of those is a decision you may one day have to defend, produced at a volume no one can reconstruct by hand. The gap between decisions made and decisions you can account for widens every quarter.

We monitor decisions, not agents.

We don’t re-decide, and we don’t grade people or systems as entities — human or machine, the questions are identical: what was known, who held authority, which rules were in force. That’s what separates an assurance record from one more dashboard watching the system.

04 What we actually deliver

When reasonable care is questioned, the record is what answers.

When a decision is challenged later, what hurts you isn’t the decision — it’s being unable to show which rules were in force, what evidence stood relative to the decision-maker at the time, and what the record shows about the steps that were taken. IDALLOS produces that contemporaneous, source-linked record.

We don’t conclude that the legal standard of reasonable care was met — that determination belongs to a regulator, a court, or the governing authority. What we make independently examinable is the evidence a reasonable-care inquiry turns on.

What we make examinable

Source-linked, time-faithful findings stating what the supplied contemporaneous record supports against the criteria then in force — evidence, authority, review, escalation, and sequence, decision by decision.

What we don’t

Rule on whether the decision was correct, or conclude that reasonable care was satisfied. Correctness and legal adequacy are for the final authority to judge — we preserve the examinable record, never a score of the decision or of ourselves.

05 How it actually works

An independent examination of decisions already made — not another system in the workflow.

A retrospective audit is an examination led by human reviewers and supported by tooling — not automated scoring. It is read-only and sits beside your workflow. You approve what governs before anything is examined; IDALLOS never assumes it.

Reconstruction follows a fixed protocol for establishing what was in force and what was known at each decision’s moment — the same discipline whether the decision-maker was a person, a model, or an agent. Who signs each record is agreed in scoping.

  1. 1

    You scope and supply

    You choose one recurring decision class and supply the fixed population of past cases and the sources that hold the evidence — records, system logs, policy and contract versions, authority matrices.

  2. 2

    We reconstruct, time-faithfully

    IDALLOS assembles the evidence and the governing criteria as they stood at each decision’s moment — the version then in force, not today’s rewrite — independent of whoever or whatever made the call.

  3. 3

    You receive the record

    A source-linked Decision Assurance Record per decision, stating what the contemporaneous record supports and — with equal prominence — what it cannot. Where the evidence is genuinely gone, we say so. We never manufacture confidence.

06 What you actually receive

The number you get — and the record behind every case in it.

Defensibility is why. This is what you receive: an independent conformance read over a declared population, and a source-linked Decision Assurance Record behind every case in it. Synthetic below, but the exact structure and discipline of the real thing.

Synthetic
Coverage Denial Authorization · 1,240 cases examined
756 / 1,240 61%

cases with all four requirements evidenced

all four evidencedone or more missing

A count over the supplied population — not a quality score. It is the AND of the four requirement rates at right.

Met, by requirement
share of 1,240 evidenced
  • Clinician review on record before denial 78% of 1,240
  • Policy version in force applied 91% of 1,240
  • Disposition within permissible set 83% of 1,240
Rule-bound consistency — comparability fixed in writing first
Consistent with comparable priors 64% of 1,240
Each rate is a count welded to its population — “Met / Not evidenced” over the 1,240 cases supplied, never a free-floating score.

One of those 1,240 records

Decision Assurance Record

Coverage Denial Authorization

DAR-CDA-0417 · examined 2026-05-12 · decision date 2024-02-08
Synthetic matter Illustrative only
Trigger

A reviewer denies an inpatient claim on the basis of a coverage limit.

Governing criteria, then in force

CP-114 rev. 3 — effective 12 Jan 2024, the version applicable on the decision date, not the current revision.

Evidence at decision time
  • Treating physician documentation present
  • Licensed-clinician review record not in supplied record
  • Clinical necessity assessment not in supplied record
  • Comparable prior determinations (2) comparability not established
Finding

On the record supplied, IDALLOS cannot establish that the required licensed-clinician review occurred before denial, or that the determination is consistent with the two comparable prior determinations.

Conclusion

The denial may have been correct. On the contemporaneous record supplied, IDALLOS cannot establish that it was supported as made.

Support by element
  • Authority to deny on coverage grounds Strong
  • Required clinician review before denial Limited
  • Consistency with comparable prior matters Limited
Support scale: Strong · Qualified · Limited — per element, no blended grade. “Cannot establish” is the finding’s conclusion, not a support level.

“A decision can be right and unsupported. Wrong and reasonable. We preserve the difference — because that difference is exactly what a regulator, a court, or a board actually weighs.”

07 Independence

You can’t be your own witness.

The system that made the decision — a person, a model, an agent, a workflow — cannot credibly vouch for itself after the fact. Self-issued confidence is not assurance. That is the entire point of an independent layer: an examiner whose only job was to reconstruct the record, who did not make the call, and who reports what the evidence will and will not support. It is also why we describe our method openly — a record you can’t scrutinize isn’t independent either.

08 The question every GC asks first

“Doesn’t this just create evidence against me?”

The exposure already exists. The decisions were made; your systems logged what they logged; a regulator or opposing party can already ask. What doesn’t exist yet is the evidence that you took reasonable care — and under a reasonable-care standard, its absence isn’t protection. It’s the thing that hurts you.

You control the posture before anything is created. A retrospective audit is a bounded exercise over a population you supply, scoped in writing in advance, with retention you set.

You decide, before any record exists
  • Engagement posture
    Engaged at the direction of your counsel, so privilege and work-product questions are decided by your lawyers before any record exists — with clear eyes that protection over a factual audit is never automatic. We structure the engagement to support the posture you choose.
    Your counsel makes the call.
  • Scope, in writing
    The decision class and population are fixed in advance. Nothing is examined or created outside it.
  • Retention, yours
    You own the output and set how long it lives. If a finding triggers a reporting or preservation duty, you learn it on your timeline, with counsel — before it’s framed by someone adverse to you.
09 How you start

One decision class. Bounded, read-only, and yours.

You don’t rebuild anything. You choose one recurring decision — adverse benefit determinations, entry filings, adverse-action notices, model-driven approvals — and IDALLOS examines a fixed population of past cases against the evidence and rules you approve. Nothing touches the live workflow.

If it finds nothing

A documented, independent basis to say this population held up to scrutiny — and a clear view of where your contemporaneous record is thin, before someone else finds it.

If it finds something

You learn it first, with counsel, and understand any resulting obligations — not just the defect — while you still have your options open.

Either way, you own it

A record built to your scope and your retention, ready if you ever need it and quiet if you don’t.

10 What we touch

Your InfoSec team’s questions, answered before the call.

You’re handing an outside examiner sensitive records, so the data posture is agreed before any engagement begins — not after.

Where the data lives

The examination can run in your environment or in a scoped, access-controlled environment we provide — your choice, fixed in the engagement.

Contracts and PII

BAA, DPA, and confidentiality terms as your regime requires. Only the sources needed for the agreed class are in scope.

Access

Least-privilege, read-only access to the supplied population. Nothing in the live workflow is touched or changed.

Return and retention

Source data is returned or destroyed per your instruction, and the records produced live only as long as you decide.

Start here

Pick the decision you’d least want to explain under oath.

That’s where we start. In a 30-minute scoping conversation we’ll name one decision class, define the population, and tell you plainly whether it can be reconstructed — no system access, no obligation. If it can’t, we’ll say so.

We’re early, and we deliberately don’t publish customer names. The first engagement is designed to produce your own proof — on your data, your decision class — not a reference you have to take on faith.