Example workflow

Compare the record. Keep judgment with the owner.

Policy comparison is a strong candidate for bounded AI because the source set can be declared, the output can be standardized, and every important difference can be checked against the record.

Concept workflow: this page describes a review pattern, not legal or compliance advice and not a claim that a particular product feature is generally available today.

Example scope: compare two approved policy versions

Job
Identify material changes between two named policy versions and prepare a source-linked change brief for a qualified reviewer.
Approved inputs
The two authoritative policy files, a defined comparison question, and any approved terminology or review standard.
Out of scope
Guessing intent, inventing requirements, using unofficial copies, deciding compliance, or publishing an updated policy.
Prepared output
A categorized change table with citations, unresolved ambiguities, possible operational impacts, and reviewer questions.
Human gate
The policy owner or qualified subject-matter reviewer validates each material change and decides what action follows.

Why a declared source set matters

A policy comparison can become unreliable when the workflow mixes drafts, unofficial copies, public commentary, or remembered requirements. The scope should identify the authoritative versions before any analysis begins.

That boundary makes omissions easier to find and prevents unrelated material from quietly shaping the result.

What a review-ready comparison should contain

A useful output is more than a summary of both documents. It should show what changed, where the change appears, why it may matter to the defined question, and what remains unclear.

  • Added, removed, and materially revised language
  • Section-level citations to both authoritative versions
  • A distinction between text changes and reviewer interpretation
  • Ambiguous or conflicting passages that require judgment
  • Operational questions for the named policy owner

Human review is the control, not a ceremonial step

The reviewer must be able to inspect the cited text, correct the categorization, add missing business context, and reject conclusions. A polished change brief is not a substitute for authority or expertise.

If the policy affects legal rights, regulated obligations, safety, employment, finance, or other high-impact decisions, appropriate professional review remains essential.

Version the scope as well as the policy

When the comparison standard changes, the workflow rules should change visibly too. Saving the scope alongside each run helps a team understand whether two briefs were produced under the same assumptions and review requirements.

A practical policy-comparison pilot

  • Confirm the authoritative old and new files with the policy owner.

  • Define the comparison question and the categories that matter.

  • Require a citation for every material-change claim.

  • Separate source text, AI interpretation, and reviewer judgment.

  • Hold ambiguous, conflicting, or high-impact items for qualified review.

  • Save the final reviewer corrections so the next scope can improve.

Useful details before you automate.

Can AI determine whether a policy is compliant?

Not by itself. AI can organize evidence and surface differences, but compliance depends on authoritative requirements, context, and qualified judgment.

Why not compare every document the team can find?

More context is not automatically better. A declared authoritative source set reduces contamination from drafts, duplicates, outdated files, and unrelated commentary.

Should every wording change be treated as material?

No. The workflow can propose categories, but a qualified reviewer should decide which changes are material to the business question.

Founding conversations are open

Bring one recurring workflow.

Describe the work, the approved evidence, and the decision a person still needs to own.

Request founding access