AMR Decision Intelligence

Trace what is funded. Verify what is executed. Identify what matters next.

AMR decisions depend on evidence distributed across national plans, surveillance systems, research, donor portfolios, projects and implementation records.

AgentBrain reconstructs those signals into a decision-grade intelligence layer. And where the chain cannot be proven, the uncertainty remains explicit.

Request Analysis Explore the Proof Stack

01Policy

02Evidence

03Capital

04Execution

05Outcomes

01Problem

The problem is not always missing data.

It is knowing what the data actually proves.

A national priority can exist without an attributable budget.

A funded health programme can support AMR without being AMR funding.

A project can disburse capital without proving implementation.

Surveillance data can exist without establishing national representativeness.

Missing public evidence does not prove missing national activity.

The individual pieces may be visible while the chain between them is not.

02From Plan to Proof

One intelligence model across the AMR decision chain.

Antibiotic susceptibility test plate with zones of inhibition — Antibiotic susceptibility evidence entering the decision chain

ClaimPriority existssource scoped

RelationEvidence supportsconfidence set

CapitalAttribution testeddirect / linked

ExecutionDelivery tracedgap retained

DecisionNext evidence rankedimpact first

Policy & Governance

What has actually been committed to? National Action Plans, priorities, responsibilities, reported capabilities and implementation structures.

Evidence & Surveillance

What is actually observed? Resistance surveillance, research evidence, coverage, representativeness and evidence quality.

Laboratory & Mechanism

What supports the biological and diagnostic picture? AST, laboratory capability, genomic evidence, resistance mechanisms and One Health signals.

Capital & Financing

What money can defensibly be attributed? Direct AMR capital is kept separate from linked programmes and wider health-financing context.

Execution & Procurement

What moved from commitment into delivery? Projects, components, procurement, disbursement and implementation evidence.

Outcomes & Impact

What can actually be connected to results? Outcomes are linked only where the evidence supports the relationship.

03Gap Discipline

Not every gap is a funding gap.

AgentBrain distinguishes between different kinds of uncertainty because the distinction changes the decision.

Attribution Gap

Capital exists, but its AMR-specific component cannot be isolated.

Classification Gap

Relevant activity exists without sufficient AMR classification.

Coverage Gap

Surveillance exists, but its representativeness remains unresolved.

Execution Gap

Funding or commitments are visible, but delivery cannot be verified.

Evidence Gap

Available information cannot confirm or reject the claim.

Outcome Gap

Activity is visible, but attributable results cannot be reconstructed.

Not visible != not present

Not attributable != not funded

Not found != absent

04Capital Intelligence

Separate health financing from AMR capital.

Large health portfolios can contain laboratory, surveillance, diagnostics, IPC, primary-care and health-security components relevant to AMR. That does not make the entire portfolio AMR funding.

Direct

Direct AMR Capital

Explicit AMR activity with defensible monetary attribution.

Linked

AMR-Linked Portfolio

AMR relevance is evidenced, but the monetary component cannot be isolated.

Enabling

AMR-Enabling Exposure

Infrastructure or capabilities may support AMR objectives without constituting attributable AMR capital.

Context

Health-Financing Context

Relevant system-level financing with no AMR attribution claim.

Unresolved

Unresolved Attribution

Evidence is insufficient to classify the capital defensibly.

05Evidence Limits

Evidence that knows its limits.

Every finding carries its own boundary. AgentBrain applies an adversarial evidence process before findings enter decision intelligence.

Scanning electron micrograph of Proteus mirabilis biofilm — Mechanism evidence remains bounded by source, scope and transformation path

Verified

Source holdsThe claim survives scope, date and provenance checks.

Inferred

Relation plausibleThe link is defensible but not equivalent to direct proof.

Candidate

Worth testingThe signal is retained without promotion into decision fact.

Unknown

Boundary explicitThe system preserves uncertainty instead of filling it.

Claim Discipline

No claim can be stronger than its underlying source and scope.

Relation Discipline

Two related records do not automatically establish a verified relationship.

Gap Discipline

Failure to find evidence is not automatically evidence of absence.

VerifiedInferredCandidateUnknown

06Active Evidence Acquisition

Beyond identifying unknowns.

Determine what is worth learning next.

Traditional analysis often ends with "more data is needed." AgentBrain asks which unresolved piece of evidence is most likely to change the decision.

Decision impact01

Information gain02

Portfolio sensitivity03

Network effects04

Resolution cost05

07Intelligence Products

Five ways to turn evidence into decisions.

Country Intelligence

Reconstruct national priorities, evidence, capital, implementation and outcomes.

Use when a country programme needs a defensible system-level view.

Capital Intelligence

Determine what funding is directly attributable, linked, enabling or unresolved.

Use when funding visibility and attribution are being confused.

Portfolio Intelligence

Map programmes, funders and interventions across countries or portfolios.

Use when additionality, overlap or duplication matters.

Execution Intelligence

Trace financing beyond commitment toward components, procurement and implementation.

Use when project totals are visible but delivery is not.

Evidence Intelligence

Identify which unresolved evidence should be acquired next.

Use when many gaps exist but only some change the decision.

08Proof Stack

Do not take the methodology on trust.

AgentBrain is being tested through multiple proof tracks. The stack includes positive results, negative findings, blocked paths and methodological limitations.

Country System Proof

Can the same architecture reconstruct a national AMR decision system?

Cross-Country Proof

Can the ontology operate across different country contexts without changing evidence rules?

Capital & Execution Proof

Can funding be separated and traced toward execution where evidence permits?

Evidence Discipline Proof

Does the system reject, downgrade or preserve uncertainty when evidence is weak?

Active Evidence Acquisition Proof

Can unresolved evidence be prioritized by modeled decision value?

Optimization Benchmark

Can alternative optimization approaches beat a strong baseline under real constraints?

Public health laboratory staff examining antimicrobial resistance isolates — Proof starts from observed records, not presentation claims

Reference casessanitized

Architecture proofstraceable

Benchmark resultscompared

Negative findingsretained

Method limitsvisible

Explore the Proof Stack

09Decision Questions

Built for decisions with real constraints.

Which AMR priorities are already financed?

Which apparent funding gaps are actually attribution gaps?

Where does capital stop being traceable to execution?

Which interventions appear genuinely additional?

Which evidence gap should be resolved before allocating more capital?

Which parts of a multi-country portfolio cannot currently be defended?

The output is a structured account of what is known, what is connected, what remains uncertain and what should be investigated next.

09AResearch Note

Africa's health financing reset.

What it means for AMR capital visibility.

Latest research11 min

Financing strategy

Why the WHO African Region health-financing reset makes AMR a useful test case.

Capital visibility

Why direct AMR capital, enabling capital and contextual capital should not be collapsed into one number.

Read research note

AgentBrain is an AMR decision-intelligence system designed around one principle: a decision system should know not only what the evidence says, but where the evidence stops.