From AI recommendation to tested decision

How Arbiter Trade Works

A decision-validation layer powered by specialised AI agents.

Accio Work
Opportunity
Arbiter Trade

Specialised perspectives

Market
Customer
Finance
Competition
Risk
Critical review
Assumptions
Tests
Test result
Learning

01 — Accio Work

The opportunity enters the system.

Accio Work · external · opportunity discovery

Smart Water Bottle — Germany

Illustrative example

Accio Work is designed to identify a product, market or commercial opportunity using AI.

Market

Germany

Launch budget

€30,000

Target price

€129

Arbiter Trade · decision validation

Is this worth committing money to?

02 — Arbiter Trade

The recommendation becomes a decision to test.

Recommendation
Assumptions
Evidence
Risks
Critical uncertainties

03 — Specialised AI agents

One decision. Specialised perspectives.

Critical Reviewer

Challenge the assumptions and identify contradictions.

Market Agent
Customer Agent
Competition Agent
Financial Agent
Risk Agent

In this method, agents analyse assumptions behind a single Arbiter Trade interface — rather than individual bots.

04 — What must be true?

The assumption to test before investing.

Illustrative example

Critical assumption

“German customers will accept a €129 price.”
Expected acceptance
62%
Evidence
Limited
Impact if wrong
High

Recommended validation

Price test — 50 respondents

Estimated cost€300

€300 tested before €30,000 committed.

Estimated duration7 days

Measured: declared purchase intent at €129. Threshold chosen for illustration.

05 — Decision record

Timestamp the decision before the test, then check it against reality.

Illustrative example

Timestamped before the test

Hypothesis: German customers will accept a €129 price.

62%

Expected acceptance

Stop threshold set in advance: below 45% acceptance, we don't launch.

Test result

36%

Outcome

Stop threshold reached: €29,700 not committed.

06 — Decision gate

Three outcomes, each tied to the evidence.

Validated

Evidence supports the decision.

Evidence: All critical assumptions supported

Validate more

Important uncertainty remains.

Evidence: 1+ assumption still ambiguous

High risk

Critical assumptions remain unsupported or contradicted.

Evidence: Price assumption contradicted

In our example, the €129 price assumption leads to High risk.

07 — The system learns from outcomes

A historical decision record, not a one-time analysis.

Prediction
Test
Test result
Prediction error
Learning
Future decisions

We want to publish how the system performs: what it anticipated correctly, where it was wrong, and where it is too early to conclude.

Company data stays private. Arbiter Trade's performance stays measurable.

Don't just ask AI what to do.
Test what must be true.

Contact us