Knowledge Partners
Universities, research institutes and technology centres advancing decision assurance, explainability and evidence-based authority.
Challenge the research →
The independent Decision Assurance application
AI agents can act. LedgerPilot determines whether they should act.
AI agents are becoming capable of acting, while enterprises still rely on human verification for consequential decisions. Google, Microsoft, Oracle and other enterprise platforms can govern agents, access, security and operations. LedgerPilot addresses the next question: is this specific business decision sufficiently supported and safe to proceed? It independently evaluates the decision, preserves the evidence in a portable Decision Assurance Record and determines the appropriate authority outcome.
Independent from the decision-maker
Portable across platforms and agents
Authority earned and reversible
Human and agentic decision sources
Human baseline
Agent decision
Workflow action
LedgerPilot Decision Assurance Application
Verify · Assure · Authorise
Zero Trust active
◎Verify
✓Assure
§Authority
AACertify
→Execute
↻Learn
Controlled enterprise execution
ERP
Business systems
Data platforms
Who should engage with LedgerPilot now?
Four audiences. One shared need: independent assurance at the point of decision.
LedgerPilot is being shaped as a platform-independent application through scientific challenge, focused co-design, enterprise validation and ecosystem integration.
Design Customers
Finance teams ready to co-design, validate and shape the first sellable D1 product in real operating conditions.
Co-design D1 →Enterprise Leaders
CFOs, CIOs and COOs exploring how AI authority can be introduced without weakening control or accountability.
Explore the business case →Technology & BC Partners
Microsoft partners, ISVs and integrators adding independent Decision Assurance to Business Central—and later to other enterprise and agent platforms.
Explore partnership →
The enterprise trust gap is measurable
AI is moving into decisions faster than trust is moving into autonomy.
A 2026 global survey of 900 CEOs by Dataiku and The Harris Poll shows the tension now shaping enterprise AI: leaders are using AI in decision-making, but human verification, explainability and legal accountability remain central. LedgerPilot is designed for that gap — at the level of the individual business decision.
still keep humans in the loop
for business-critical AI decisions. Capability is advancing, but enterprises are not yet ready to grant unconditional autonomy.
ask AI to justify recommendations
at least occasionally. Trust is conditional: consequential decisions need evidence, explanation and a defensible basis.
are concerned about legal risk from AI agents
as agents take on more responsibility. The question is no longer only what an agent can do, but what the enterprise can responsibly allow it to do.
This is the space LedgerPilot addresses.
Agent platforms govern capability and technical permission. LedgerPilot independently determines whether a specific enterprise decision is sufficiently supported, explainable and assured to proceed.
Source: Dataiku / The Harris Poll, Global AI Confessions Report: CEO Edition 2026. Survey conducted February–March 2026 among 900 CEOs of companies with annual revenue of at least US$500 million or regional equivalent. Statistics are paraphrased from the published survey findings.
The Zero Trust starting principle
Permission to act is not proof that a decision is sound. LedgerPilot determines whether it has earned the right to proceed.
Is the decision sufficiently assured?
LedgerPilot independently evaluates evidence, provenance, decision quality, policy compliance, risk and explainability before the decision may proceed.
Is the agent's identity and performance verified?
The agent is evaluated by decision class, operating context, observed outcomes, exceptions and human interventions—not by generic capability claims.
Is the agent authorised for this decision?
Assurance evidence and the Delegation of Authority envelope determine whether the agent may advise, propose, decide or execute this specific decision.
Under which guardrails may it proceed?
Every AA Certificate is scoped, monitored and reversible, with human escalation, override and revocation embedded in the operating model.
The product position
One independent application. Four decision-level capabilities.
Generic governance platforms manage the environment around AI: agents, identity, access, security, orchestration and monitoring. LedgerPilot does not duplicate those capabilities. It integrates with them and independently evaluates the individual business decision—separating the decision-maker from the decision-assurer and keeping the assurance result portable across platforms.
Layer 01 · Zero Trust evidence foundation
Decision Evidence & Integrity
Every decision starts untrusted. LedgerPilot defines the decision boundary, captures the relevant evidence and verifies its source, completeness, context and integrity before independent assurance begins.
Human Decision Capture Layer
Agent identity
Decision context
Evidence provenance
Outcome evidence
Standardised decision record
The decision—not the whole platform—is the product focus
Assure the decision. Verify the authority. Certify whether it may proceed.
The AA result is not a generic platform safety label or a blanket promise of zero risk. It is decision-specific evidence that the applicable assurance and authority conditions were met within an explicit, monitored and reversible scope.
Verify
Confirm the decision source, agent identity, context, provenance and completeness of the evidence before any trust is granted.
Assure the decision
Use EDA to evaluate decision quality, policy compliance, confidence, risk, provenance and explainability.
Verify the agent
Measure identity, performance and reliability by decision class using accuracy, consistency, exceptions, interventions and realised outcomes.
Verify authority
Use AGA to confirm that the agent holds the required Delegation of Authority for this decision, context and risk level.
Issue the AA Certificate
Certify the decision as Assured and Authorised only when the applicable evidence, policy, risk and authority conditions are met.
Execute, observe & learn
Execute within guardrails, record the outcome and use new evidence to maintain, expand, reduce or revoke authority.
The first proving ground · D1
Begin with one bounded decision. Prove independent assurance in enterprise reality.
LedgerPilot begins with Supplier Invoice Posting Eligibility in Microsoft Dynamics 365 Business Central. D1 independently evaluates whether a proposed posting decision is sufficiently supported, assigns an assurance grade and produces an auditable Decision Assurance Record. It starts read-only and advisory-first, with humans accountable. Business Central is the first evidence environment—not the boundary of the product.
100% human decisions is a valid—and valuable—starting point.
AA Certified by LedgerPilot
Assured and Authorised within defined scope, backed by a Decision Assurance Record.
Capture current AP decisions
Make exceptions, approvals, posting choices and human interventions visible without changing who decides.
Replay and shadow
Reconstruct historical decisions, then score live cases in parallel to operations—with no autonomous execution.
Earn bounded authority
Introduce advice first, then grant narrow decision authority only to proven decision classes within explicit limits.
The decision-authority maturity model
No trust by default. Authority is earned—not assumed.
Starting point: 100% human decisions and 0% autonomous agent authority.
Historical replay
Reconstruct
Establish the human decision baseline, formalise decision classes and calibrate the first trustworthiness measures.
Live shadow
Observe
Evaluate live decisions and agent performance in shadow mode while all formal authority remains with humans.
Advisory
Recommend
Agents recommend actions; LedgerPilot values their decisions and performance while accountable humans retain the DoA.
Bounded autonomy
Delegate
AGA transfers narrow DoA for proven decision classes within explicit limits, continuous monitoring and immediate revocation controls.
Governed expansion
Scale safely
Sustained performance supports broader authority, while continuous assurance can maintain, reduce or revoke it as conditions change.
Toward a VLAIO research project
What would it take to make AA certification of agentic decisions scientifically defensible?
LedgerPilot is being shaped as an ambitious research-driven Decision Assurance application in Flanders. It starts from a Zero Trust thesis, a platform-independent product boundary and a practical first proving ground—but its decisive questions cannot be resolved through conventional software development alone. The next step is to bring together the scientific, architectural and industrial capabilities required to formalise the assurance science, engineer the authority controls and validate decision-level assurance in enterprise reality.
For knowledge partners, design customers and technology partners
We invite the right partners to challenge the thesis, shape D1 and help establish the evidence.