Inside the engine

Aurum

The autonomous research engine behind every report. It searches, verifies, challenges its own conclusions, and delivers decision-ready intelligence — before you step in.

The process

Four stages. No human shortcuts.

Every report follows the same sequence. The depth of each stage scales with the question, but the structure never changes.

1
Search
Aurum breaks your question into sub-questions and searches across institutional databases, government publications, industry reports, news archives, and market data. Each source is retrieved live — not recalled from training data. The goal is coverage: find everything relevant, miss nothing critical.
2
Verify
Every source is graded A, B, or C based on provenance, recency, and institutional reliability. Claims are cross-referenced against multiple independent sources. Contradictions are flagged, not hidden. The evidence base is built before any conclusions are drawn.
3
Challenge
A separate adversarial review evaluates the draft: scoring evidence quality, testing logical coherence, surfacing counterarguments, and identifying weak points. If the score falls below threshold, the report is sent back for improvement. This cycle repeats until the conclusions survive scrutiny.
4
Deliver
The final output: a bottom line with calibrated confidence, probability-scored scenarios, counterarguments per finding, recommended actions with time horizons, and the full evidence report with graded source citations. Decision-ready, not information-heavy.
Intelligence modes

12 research templates. Automatically selected.

Aurum analyzes your question and selects the research template that fits. Each mode shapes the search strategy, source priorities, analytical framework, and output structure.

Scenario Analysis
Maps possible futures with probability scores, triggers, and indicators. For questions about what could happen and how likely.
Due Diligence
Systematic risk and opportunity assessment. Financial, legal, operational, and reputational dimensions before capital is committed.
Market Research
Market sizing, segmentation, demand signals, and competitive landscape. Quantitative where data exists, calibrated where it does not.
Competitor Audit
Positioning, strategy, capabilities, and vulnerabilities of specific actors. Based on public filings, hiring patterns, product signals, and market data.
Investment Thesis
Tests the logic of an investment case. Bull and bear arguments, key assumptions, what would need to be true, and what could break the thesis.
Risk Assessment
Identifies, categorizes, and scores risks by likelihood and impact. Regulatory, operational, geopolitical, and market risk in one view.
Strategic Outlook
Medium-term strategic analysis: what is changing, what it means for the sector, and what the smart positioning looks like over 12–36 months.
Market Watch
Current-state monitoring. What has changed recently, what signals matter, and what requires attention now. Time-sensitive by design.
Deep Research
Open-ended, exhaustive investigation of a complex topic. Maximum source depth, multiple analytical lenses, full evidence synthesis.
Actor Mapping
Identifies key players, their relationships, influence networks, and strategic incentives. Who matters, who moves first, who follows.
Scientific Review
Evaluates scientific and technical evidence. Study quality, consensus status, methodological concerns, and practical implications.
Political Risk
Assesses regulatory, legislative, and geopolitical risks to specific business decisions. Election cycles, policy trajectories, enforcement patterns.
The agent selects the mode automatically based on your question. No configuration required.
Two depths

Same engine. Different intensity.

Every report uses the full pipeline. The difference is how many rounds of research, how many sources, and how many times the adversarial review pushes back.

Decision Brief — €249
SourcesUp to 90 verified
Research rounds1–2
Adversarial passes1
Delivery~7–25 min
OutputBrief + PDF + full report
Follow-up
Deep Intelligence — €499
SourcesUp to 400 verified
Research rounds3–5
Adversarial passes3–5
Delivery~20–90 min
OutputBrief + PDF + full report (10,000+ words)
Follow-upDecision Room + challenge findings
Source grading

Not all sources are equal. Aurum knows the difference.

Every source is graded before it enters the evidence base. The grade affects how much weight it carries in the analysis and how it is cited in the final report.

A
Primary / institutional sources
Government publications, regulatory filings, court documents, central bank data, peer-reviewed research, official statistics, company filings (10-K, annual reports). Direct, verifiable, authoritative. Highest evidential weight.
B
Secondary / established sources
Major news organizations, industry analyst reports, established trade publications, named expert commentary, think tank research. Reliable but one step removed from primary data. Claims are cross-checked against A-grade sources where possible.
C
Market signals / supplementary
Prediction markets, social media signals, blog posts from domain experts, forums, startup databases, job postings, patent filings. Useful for directional signals and emerging trends. Never used as sole evidence for a key finding.
Adversarial review

Every conclusion is attacked before you see it.

After the research agent builds its case, a separate evaluation process scores the work and decides whether it is good enough to deliver. This is not a simple check — it is structured criticism.

01
Evidence audit. Are the claims supported by the cited sources? Are the sources graded correctly? Are there claims that rest on a single source?
02
Logic check. Does the reasoning hold? Are there leaps in the argument? Does the bottom line follow from the findings, or does it overreach?
03
Counterargument surfacing. What is the strongest case against the conclusion? Is it addressed in the report? If not, it must be.
04
Scoring. The review produces a quality score. If below threshold, the report is sent back to the research agent with specific improvement instructions.
05
Iteration. The research agent addresses the weaknesses, gathers additional evidence if needed, and resubmits. The cycle repeats until the review passes. Deep Intelligence reports go through 3–5 passes.
Confidence methodology

Probabilities, not opinions. Calibrated language.

Aurum uses the Sherman Kent / NATO standard for expressing uncertainty. Every conclusion carries a probability estimate with defined meaning — not vague hedging.

Term
Probability
Meaning
Almost certain
93%+
The evidence is overwhelming and consistent. Extremely unlikely to be wrong.
Probable
75–92%
Strong evidence with minor gaps. The conclusion is well-supported but not beyond question.
Chances about even
40–74%
Evidence supports the conclusion but significant uncertainty remains. Competing interpretations are plausible.
Probably not
15–39%
The evidence leans against this outcome, but it cannot be ruled out.
Almost certainly not
<15%
The evidence strongly contradicts this outcome. Would require extraordinary developments.

This framework ensures that when Aurum says "probable," it means 75–92% confidence based on the available evidence — not a feeling. Every finding in every report carries this calibration.

See the depth for yourself.

Order a report. Ask any high-stakes question. Get the evidence, the scenarios, and the call.

Order a report →