← The Decision Architecture

02 · ARCHITECT · MONARC AI SYSTEMS DIAGNOSTIC™

What your systems show, not just what we are told.

A deeper, read-only look at the AI systems and agents behind the decisions in your design. It is scoped inside The Decision Architecture, under its own written authorization, when the design needs evidence from your own systems.

  • Read-only
  • In your own tenant
  • Settings and metadata only
  • A person confirms every result

01Where it fits

Two looks at your AI, at two stages.

The Assessment maps your AI with your leadership team. The Diagnostic comes later, only if the design needs it, and reads what your systems themselves record.

ASSESS · INCLUDED IN THE ASSESSMENT

Guided AI Inventory

Source
Your team, with the Monarc Principal Architect
Answers
Where AI touches your consequential decisions, and who owns it
Evidence
Mostly what people report and what your documents show

ARCHITECT · SCOPED SEPARATELY

Monarc AI Systems Diagnostic™

Source
Read-only settings and metadata from your own tools
Answers
Whether the controls the design relies on are in place as described
Evidence
What can be observed in your systems, on a stated date and scope

02How it works

Authorized first, then read.

  1. 01

    Authorize in writing

    A separate authorization names each source, what is read, who runs each export and how long anything is kept. Agreed security controls are in place before Monarc receives any file.

  2. 02

    Read, without changing anything

    Your team runs read-only exports in your own tenant. They collect settings and metadata, never the contents of emails, files or chats. Each export is fingerprinted so its origin can be traced. It works with the platforms you already run, whether Microsoft 365, Google Workspace or another suite, through your identity provider and the admin consoles of the AI tools you use. Where a platform does not show something, the result says so.

  3. 03

    Match each system to its owner

    Every AI system and agent found is matched to the decisions it touches and the person who owns it. AI may help sort entries into categories; it never confirms one.

  4. 04

    Check against the design

    Each control the design relies on is checked against criteria agreed in scope. Every result states what was expected, what was found, and the evidence behind it. Where something cannot be seen, the result says so.

  5. 05

    Confirm with your owners

    The Monarc Principal Architect reviews every result, and your owners confirm the facts before anything is final.

03What it produces

Evidence your design can stand on.

01 · REGISTER

AI systems register

Each AI system and agent with its owner, the model and version it runs, its vendor, where it runs and who can stop it.

02 · FINDINGS

Findings with evidence

What was expected, what was found, and how it is known: reported, documented or observed.

03 · OWNERS

Blockers first, by owner

Anything that stops a decision from holding is grouped under the person who owns that decision.

04 · TRACE

A fix path for every finding

Each finding carries through requirement, control design, work package, test and acceptance in your design.

05 · PACK

Dated evidence pack

A fingerprinted record of what was read and found, kept in your own workspace and shared only with your consent.

06 · DESIGN

Inputs to your sign-off

The results feed the Authority Matrix, Enforcement Layer and Build Backlog you approve at the end of the architecture.

Diagnostic record · illustrativeRead-only · your tenant

Register entry

SYSTEM
Invoice approval agent
DECISION
Supplier payments under a set limit
OWNER
Controller
MODEL
Vendor model, version recorded
RUNS IN
Your own tenant
CAN STOP IT
Finance systems lead

Finding · stop control

OBSERVED
EXPECTEDA named person can pause the agent, and the pause is recorded.
FOUNDPause is possible, but no record of who paused it or when.

Owner confirmed the facts. A fix path opens in the design.

Requirement

WHAT MUST BE TRUEEvery pause is recorded with who, when and why, and the owner is told.
WHYThe design relies on a stop that can be shown to have worked.

Agreed with the Controller, who owns the decision.

Control design

HOW THE DESIGN MEETS ITPausing goes through one recorded step, by the named person or their backup.
WHERE IT SITSIn the Enforcement Layer of your design, beside the decision it protects.

Part of the design you sign off.

Work package

WHAT WOULD BE BUILTThe recorded pause step and a notice to the owner, in the tools you already use.
WHOYour Finance systems lead, with Monarc if you approve a build.

Nothing is built at this stage. Building is a separate decision.

Test

HOW IT IS SHOWN TO WORKThe agent is paused in an agreed test window, and the record is checked.
PASSES WHENThe record shows who paused it, when and why, and the owner was told.

The test is written before anything is built.

Acceptance

WHO SIGNS IT OFFThe Controller accepts the test result.
THENThe finding closes, with the date and the evidence kept in your workspace.

Acceptance belongs to you, not to Monarc.

Tap each step to follow one finding to its fix.

Illustrative example, not a client's record.

04Your data and your choices

Scope before access. Nothing changes in your systems.

What we hold to

  • Read-only. The Diagnostic changes no setting in any system.
  • Your data stays in your own tenant and workspace. Monarc does not host your decision data.
  • Settings and metadata only, never the contents of your emails, files or chats.
  • No file is received before the written authorization and agreed security controls are in place.
  • You can ask us to keep AI off your material entirely.

What it is not

  • Not a score. The AI Decision Score comes only from the Assessment, set by the Monarc Principal Architect.
  • Not an audit, a certification or a penetration test.
  • Not a build. Fixes are designed here and built only if you approve a separate build.
  • Not legal advice. We work alongside your counsel, never in their place.

Monarc also architects decision systems. The AI Decision Score is assigned before any remediation scope is discussed.

Start with the Assessment.

The Diagnostic is available to organizations that have completed the AI Decision Assessment and authorized The Decision Architecture. Scope, sources and fee are set in writing for each engagement.

About this page. Engagement descriptions are summaries. Scope, sources, deliverables and fees are confirmed in writing.

Not an audit or certification. Monarc assesses, architects and builds. It does not audit, certify, attest or provide legal, insurance or regulatory advice.