Ask your workplace's contract and policies a plain-English question — then check every part of the answer, quote by quote, against the actual source. The AI helps you find it; the program proves it's really there.
It opens loaded with a small sample library. Ask something like "How much notice do I need to give to resign?", "How am I evaluated?", or "How does longevity pay work?" — the tool finds the right document, and after you paste an AI's answer it marks each claim Verified, Review Needed, or Not found. No sign-up, nothing leaves your browser.
Rules are only equally available when people have an equal ability to find and understand them.
In many workplaces, the documents governing employment are technically available to everyone: contracts, board policies, employee handbooks, salary guides, procedures, and other written rules.
Access to the documents, however, does not create equal access to the information inside them.
An organization may have administrators, attorneys, human-resources staff, institutional knowledge, and working time available to research those documents. An individual employee may have a PDF, a question, and whatever time they can find to search hundreds of pages for an answer.
Governed Evidence Validator (GEV) was built to help narrow that gap.
GEV allows a person to ask a plain-English question, use the AI of their choice to help analyze the relevant documents, and then bring that analysis back to a separate deterministic system that checks the claims against the actual source material.
The AI can help make the information accessible.
The documents remain authoritative.
And the person asking the question gets to see the evidence for themselves.
The rules governing a workplace can be public, available, and still be practically inaccessible.
Knowing that an answer exists somewhere within a collective bargaining agreement, employee handbook, or hundreds of individual policies is very different from having the time and expertise to find it.
That creates an information imbalance.
The party administering those rules may work with them every day and have professional resources available to interpret them. The person governed by those rules may encounter them only when something has already gone wrong: a leave question, a resignation, a grievance, a disciplinary issue, or a disagreement over procedure.
AI creates an opportunity to reduce that imbalance because it can make large collections of difficult documents much easier to explore.
But simply replacing one information disadvantage with an unverified AI answer creates a different problem.
A fluent answer is not evidence.
GEV was designed around that tension.
It uses AI for what AI can do well—helping people navigate, synthesize, and understand complex information—while keeping evidentiary authority outside the model.
The objective is not to tell someone what their contract means.
It is to make it easier for them to find the language that governs their situation, understand the analysis being offered, and inspect the evidence before deciding what to do with it.
The LLM proposes. The program verifies. The program decides.
GEV separates assistance from authority.
The language model can help a user investigate documents that might otherwise require hours of manual searching. But the model does not get to declare its own analysis trustworthy.
The validator independently checks the evidence.
The source documents remain visible.
The human remains the final judge of whether the evidence actually supports the conclusion.
This separation allows AI to reduce the cost of accessing information without requiring the person using it to surrender judgment to the AI.
Information asymmetry has practical consequences.
A rule buried in a lengthy contract or policy manual may technically be available to everyone while remaining functionally easier for one side to use than the other.
GEV cannot eliminate differences in legal expertise, institutional knowledge, representation, or resources.
It can reduce one smaller but meaningful part of that imbalance: the time and difficulty involved in finding the governing language and checking whether an explanation is actually grounded in it.
That is why the system exposes its evidence rather than simply returning an answer.
The goal is not:
"Trust this program instead."
The goal is:
"Here is the source. Here is what the AI said. Here is what the validator found. Now you can check."
GEV demonstrates the validation principle of the Operational Integrity System (OIS): model output should be independently verifiable against an authoritative outside source.
Within GEV, that principle is applied to document analysis. The language model remains free to reason and explain, while a separate deterministic process evaluates whether the evidence claimed in that analysis can actually be traced to the documents provided.
OIS provides the broader governance framework.
GEV demonstrates what one part of that framework can look like when applied to a practical problem.
Governed Evidence Validator is a tool for making AI-assisted document research more accessible and inspectable.
It combines plain-language questioning, user-selected AI analysis, deterministic evidence validation, and direct access to the supporting source language.
It was designed for situations where the person asking the question may not have the same time, familiarity, or resources as the institution responsible for administering the rules.
GEV is not a lawyer, union representative, legal interpretation engine, or substitute for professional advice.
It does not determine what a contract or policy legally means, decide whether an employer or employee is correct, or guarantee that an AI-generated interpretation is sound.
It verifies traceability.
The system helps a user determine whether the evidence an AI relies upon can actually be found in the governing documents and then puts that evidence in front of the user for review.
The documents governing people's working lives should not become useful only when someone has the time, expertise, or institutional resources to search them effectively.
AI can help make those documents easier to explore. But greater accessibility should not require greater trust in an opaque system.
GEV was built around a different tradeoff.
Use AI to make the information easier to reach. Use deterministic software to check the evidence. Keep the source visible. Keep the person in control.
If GEV helps someone walk into a conversation, meeting, or decision with a better understanding of what their own governing documents actually say, then it has accomplished what it was built to do.