Below the line of visibility

Everyone is watching the front stage. We built the backstage.

AI got better at answers. OmnIA records what each answer stood on: the events, the evidence and the approvals behind every decision, in order.

For firms that run on judgment: law, consulting, family offices and others. The system connects and proposes. The people with authority decide.

Stage of the solution

The value proposition is complete. The product is in validation, and it will change with what validation shows.

Concept done

The thesis and the method. Documents organized and classified by entity.

Validate ← we are here

The chain of events behind a real decision. The human approval gate. Answers that say “not on record”.

Optimize

Recording as a side effect of daily work. Records from many agents in one place.

Replicate

A second domain. Other kinds of firms.

stages follow the Design Doing Harness · each stage closes on a human gate

The problem

Firms keep documents. They lose the reasoning.

Every serious decision draws on three things: the documents on file, what the organization learned before, and the judgment of the people in the room. The documents stay. The other two leave with the people.

When a partner moves on, the reasons go with them. When a decision is questioned five years later, nobody can show what it stood on, who weighed it, or what was known at the time.

Documents
History
Judgment

Why now

direction · optimize stage

Everyone will have an agent. Who keeps what they learned?

Agents read, weigh and propose all day, and the reasoning is gone when the session ends. Each person gains an assistant. The organization gains nothing it can keep.

OmnIA is the shared record: one place where what people and their agents worked out is kept with its evidence, its author and its approval.

How it works

validating

Ask why, and the path back is already there.

Work moves forward: research, findings, insight, recommendation. The question arrives the other way: why did we decide this? Without a record, someone rebuilds the path by hand, if anyone still can. With OmnIA, the path back exists before the question.

It matters more now. AI proposes without showing how it got there. OmnIA records how each piece of knowledge was produced: by which person or agent, from which sources, and who accepted it.

Without OmnIAfrom the front stage, both look the same
A decision without OmnIA The client sees a question and a recommendation. Behind the line of visibility, the material lies in scattered fragments: interview notes on a laptop, a market report in a folder, a closed AI session, a photo of post-its, and a rejected option nobody wrote down. Years later there is no path back. FRONT STAGE LINE OF VISIBILITY BACKSTAGE WHAT GETS KEPT Question where should we grow? Recommendation launch service X Interview notes one consultant’s laptop Market report shared folder, v7 AI session closed, not saved Workshop synthesis photo of post-its Rejected option never written down ? years later, “why this?” has no path back · the answer depends on who is still around
With OmnIAvalidating
The same decision with OmnIA The same question and recommendation. Behind the line, interviews and a market report feed coded findings. Two readings compete: reading A, proposed by an agent and accepted by a manager, is chosen; reading B is rejected with its reason kept. Interviews with segment B are flagged as missing evidence. The decision record is signed by a partner. Asking why traces the path back from the recommendation to its sources. FRONT STAGE LINE OF VISIBILITY BACKSTAGE OMNIA CHOSEN REJECTED · REASON KEPT “WHY THIS?” · TRACED BACK FROM THE DECISION TO ITS SOURCES Question where should we grow? Recommendation launch service X 12 interviews primary · testimony Market report secondary · document Findings coded · page refs Reading A agent · accepted J.R. Reading B one source only Decision record signed · partner Segment B interviews missing evidence each block: who or which agent made it · from which sources · who accepted it · append-only
illustrative · no client datasolid: strong evidence · dashed: single source or missing
Without OmnIA
With OmnIA
Sources
scattered across people and tools
linked to the case, with their origin
Alternatives
forgotten
kept, with the reason
Gaps
invisible
flagged as missing evidence
AI output
accepted without a trail
marked as proposed, until a person accepts it
Approval
implicit
signed, at each person’s level
Years later
“ask whoever was there”
the decision rebuilt as it stood that day
Documentmaterial fact

Stands on its own

A contract, a notice, a title deed. The strongest kind of evidence.

Testimonywhat someone said

Weighed by its source

An email, a voice message, a meeting note. Needs corroboration from other evidence.

Interpretationan analysis

Never evidence of itself

A reading of the case, including one made with AI. It has to rest on documents and testimony.

interpretations cannot cite themselves as evidence · every factual answer points to document, page and hash

Many hands

in design

Many hands, one record.

In a consultancy, the same question is worked in separate streams: one team on the numbers, another on interviews, a third on regulation. OmnIA links what each stream produced, shows where the readings disagree, and takes the choice to whoever has the authority to make it.

Each approval is recorded at its level. An analyst confirms a source. A manager accepts a reading. A partner signs the recommendation.

decision rights are part of the record · who may approve what is defined before the system goes live

What it shows

validating

A broken link is visible.

A decision with no document behind it. An approval with no author. A case where nothing has been decided. OmnIA shows the gap instead of filling it.

Decided

With the evidence it stood on, who proposed it and who approved it.

Missing evidence

A step with nothing underneath it. Shown as missing, never filled with a guess.

Open

Cases where no decision exists yet, listed as open.

empty retrieval returns “not on record” · it never generates a source

For law firms

in development

Chain of custody for the reasoning behind a case.

A complex case rests on claims, evidence, precedents and answers to objections. OmnIA keeps that chain and shows where a claim has nothing underneath it, where an objection was never answered, where a key document is missing.

It does not judge the merits. It shows the structure, so your team can. The same applies to a consultancy’s recommendation, or a family office decision resting on an assumption no one wrote down.

The method

A method first. Then the system.

Every deployment starts with your people, not with software. The system is set up only after the flows and the rules are agreed.

01

Understand

How information moves through the business, observed.

02

Divide

What belongs to the machine and what belongs to people.

03

Design

Flows, access and governance.

04

Approve

Nothing goes live without your sign-off.

05

Deploy

The system runs, with the human gate built in.

OmnIA records what passes through it. Nothing else. We apply to our own work the rule the system enforces on yours.

Under the hood

The system proposes. The people with authority decide.

  • Built with Claude. The answering layer moves to Claude during validation. validating
  • Every change to OmnIA is checked by an independent reviewer from a different model family before a person approves it. built
  • Approvals are recorded as signed events, with the reason. validating
  • The log is append-only and hash-chained. The past cannot be rewritten in silence. validating
  • Client documents stay on infrastructure the client controls. What may reach a language model is agreed with the client. built
event-sourced: the state of a case is derived from its log · replay any decision as it stood on a given date

The thesis

A system of record for decisions.

Git gave software a memory of every change and why. OmnIA does the same for decisions.

Git
OmnIA
Git
OmnIA
diff
evidence
commit message
the reason
author
who proposed
parent commit
prior decision
branch
an option considered
merge
the option chosen
kept branches
options rejected, and why
no equivalent
the broken link

Companies have systems of record for money, customers and code. None for decisions.

The method has a lineage: contribution analysis (Mayne), realist evaluation (Pawson and Tilley), and a four-layer evidence protocol (existence, quality, internal relation, strength) first used in a multi-product evaluation for an international organization. The core is the same in every field. What changes from a holding to a law firm to a clinic is the sources it reads.

Cases

solid: concluded · dashed: in progress

Where it is being tested.

ConcludedPublic sector

Where the method comes from

A results evaluation for an international organization: sources, analysis, alternative explanations, and the one the evidence supported.

ValidatingAgribusiness

Family holding

Land titles, contracts and licenses organized by property. Now recording the chain behind a real decision.

NextHealthcare

Occupational health clinic

The second domain, to test that the core holds outside the first one.

Who is behind it

Leandro Pasini

Design, service design and public innovation for more than twenty years, on a base in programming that goes back to the 1980s. Co-author of the Design Doing method. In innovation consulting he rebuilt by hand the path from each recommendation back to its sources. OmnIA makes that path exist before anyone asks.

Contact

Talk to us.

Tell us what kind of decisions your firm makes and where the reasoning gets lost.

opens your email app · contato@omniaos.com.br