AI Research Analyst

Verified Intelligence for Your Domain

A structured AI research methodology that eliminates hallucination and fabricated sources. Your team gets domain-specific research guides with built-in source integrity enforcement - every claim traced, every source verifiable, every inference labeled.

Book a 30-Minute Discovery Call See How It Works

Your AI Is Making Things Up. You Just Haven't Caught It Yet.

Large language models fabricate sources, invent statistics, and present fiction as fact - with complete confidence. For teams whose work depends on accurate research, this is not an inconvenience. It is a liability.

  • Compliance teams cite AI-generated regulatory references that do not exist
  • Due diligence reports contain fabricated company data that no one verifies until closing
  • Grant applications reference hallucinated impact studies that reviewers flag on page one
  • ESG reports built on AI output fail audit because the underlying sources are invented
  • Medical device submissions include material properties sourced from non-existent papers
  • Every fabricated citation that reaches a client, regulator, or investor is a reputational event

Visual Element

The Zbigniew Protocol: Source Integrity by Design

Named after Zbigniew Herbert - the poet who demanded truth without decoration. The protocol enforces source verification at every step of AI-assisted research. Fabricated citations are caught by design, not by luck.

1

SOURCE COMPLIANCE Tally - Every output includes a running count: sources cited, sources verified, sources pending, sources rejected. Your team sees the integrity score before acting on any finding.

2

Mandatory Source Classification - Every claim is tagged: PRIMARY (peer-reviewed, official), SECONDARY (industry report, news), TERTIARY (blog, opinion), or INFERENCE (analyst reasoning). No untagged claims pass through.

3

Fabrication Detection Rules - The protocol includes explicit instructions that prevent the AI from generating plausible-sounding but non-existent references. When the model cannot find a real source, it says so - instead of inventing one.

4

Domain-Specific Verification Chains - Each domain guide includes the authoritative source hierarchy for that field. Medical device research checks against FDA/MDR databases. AML checks against FATF and local FIU registers. No generic research patterns.

5

Audit-Ready Output Format - Every research output is structured for handoff: executive summary, detailed findings with inline citations, source appendix with verification status, and confidence assessment. Ready for compliance review without reformatting.

Who This Is For

Teams where 'the AI said so' is not an acceptable source citation.

Compliance Teams

Drowning in regulatory research across multiple jurisdictions. Need audit-grade source trails for every finding. Cannot afford a single fabricated reference in a submission.

Due Diligence Analysts

Running background checks, market analysis, and risk assessments where every data point must trace back to a verifiable source. The cost of a hallucinated fact is a failed deal or a lawsuit.

Grant Writers

Building impact pathways and citing research to justify funding requests. Reviewers check sources. Hallucinated impact studies get your application rejected - and your organization flagged.

Quality & Regulatory Teams

Medical devices, pharma, manufacturing. Material properties, biocompatibility data, and regulatory precedents must come from real publications. No room for AI-generated references.

ESG Reporting Teams

Preparing for CSRD with evidence-based reporting. Auditors will verify your environmental claims and social impact data. Every number needs a traceable source.

Strategy & Research Departments

Using AI to accelerate market research, competitive analysis, and trend identification. Need a methodology that separates verified intelligence from confident-sounding fabrication.

Three Ways In

Start with the open-source protocol. Go deeper when your domain demands it.

CORE - Free

The Zbigniew Protocol on GitHub. Open-source. Self-service. Download the CORE.md research guide and start using it with ChatGPT, Copilot, or Claude today. Source compliance tally, fabrication detection rules, and audit-ready output format included. github.com/maciejjankowski/zbigniew

DOMAIN PACK - 1,500 EUR

Custom domain guide built for your field, plus a 60-minute onboarding call showing your team exactly how to use it. Available domains: Supply Chain Continuity, Medical Device Materials, AML/Due Diligence, ESG Reporting, EU Grant Research. Includes CORE.md, domain-specific guide, onboarding session, and 30 days of email support.

ENTERPRISE - 5,000-15,000 EUR

We analyze your research workflows, build a tailored guide for YOUR industry, train your team, and set up source verification processes. Includes everything in Domain Pack plus custom guide development, half-day on-site or remote training, source library setup, and quarterly review.

Your AI Will Not Fabricate Sources

The Zbigniew Protocol is not a prompt template. It is a structured methodology with enforcement mechanisms built into every research output.

  • Every output includes a SOURCE COMPLIANCE tally - fabricated citations are protocol violations caught by design
  • Claims without verifiable sources are labeled INFERENCE, not presented as fact
  • Domain guides include the authoritative source hierarchy for each field - the AI knows where to look
  • The protocol is open-source - inspect every rule, adapt to your workflows, verify it works before you pay
  • Onboarding shows your team how to use it with the AI tools they already have - no new software to buy
  • Enterprise engagements include source library setup so your team's verified sources compound over time

Visual Element

Start With the Free Protocol. See What Your AI Has Been Making Up.

Download the Zbigniew Protocol from GitHub and run it on your last research output. If the SOURCE COMPLIANCE tally makes you uncomfortable, book a call to discuss a domain-specific guide for your team.

Book a 30-Minute Discovery Call