AI for HR

When Two Approved Policies Apply, AI Should Not Decide in Silence

ShareLinkedIn

SimplSolutions Editorial Team · 6 min read

Published

When Two Approved Policies Apply, AI Should Not Decide in Silence - featured article image

The difficult part is not finding a policy

An employee asks: “My manager approved a different schedule for me. Does that approval override the standard attendance policy?”

An AI assistant may find two relevant, approved sources:

  • The standard attendance policy, which sets the organization-wide expectation.
  • A manager-approved schedule exception, which appears to authorize a different arrangement for this employee.

Both sources can be current. Both can be legitimate. The hard question is not retrieval. It is authority: which rule governs this employee, in this situation, for this period?

That distinction matters because a confident answer can become an instruction. If the assistant says the exception applies when it does not—or says the standard policy controls when a valid exception does apply—the employee may act on an answer that no authorized person actually approved.

Separate relevance from authority

A useful HR-owned approach treats the answer as two different judgments.

Relevance asks: Which approved sources relate to the question?

Authority asks: Which source governs this employee and circumstance?

An assistant can help with the first judgment. It can locate the attendance policy, the exception record, and related scope information. But relevance alone does not establish priority. A manager approval may be limited by role, duration, business unit, accommodation process, collective agreement, or another approval requirement. The standard policy may also contain an exception rule that is easy to miss.

The assistant should therefore make the overlap explicit rather than compressing it into a single answer. A responsible response might say:

“I found the standard attendance policy and a manager-approved schedule exception that may apply. The available information does not establish whether this approval overrides the standard policy for your situation. HR policy ownership is needed before you rely on one interpretation. I’m routing this for review.”

That is not a failure to answer. It is an accurate representation of what is known and what remains unresolved.

Worked example: the schedule exception

Consider an employee whose standard schedule requires attendance on particular days. The employee’s manager has approved a different schedule for the next month. The employee asks whether missing a standard attendance day will count as an attendance violation.

The assistant should identify, at minimum:

  1. The governing subject: attendance, scheduling, or both.
  2. The employee scope: whether the policy and exception apply to this worker, location, employment category, or team.
  3. The authority of the approver: whether the manager is authorized to create or approve this type of exception.
  4. The effective dates: when the standard policy and exception apply, and whether either has expired.
  5. The exception terms: whether the approval changes the attendance requirement, changes only the schedule, or leaves attendance consequences untouched.
  6. Any higher-order rule: whether another approved policy or process governs the question.

The assistant can summarize those facts without deciding the outcome. It can also identify what is missing—for example, the exception’s end date or the approval authority required by the attendance policy.

If the documents clearly establish that the manager may approve this exception, that the exception covers the employee and dates in question, and that it changes the relevant schedule requirement, the assistant may present that documented conclusion with citations to the approved sources. If those conditions are not clear, it should pause and route the question to the designated HR policy owner.

The key control is not a particular phrase. It is the refusal to turn an unresolved authority question into a silent decision.

Use a decision state, not just an answer

HR teams should define a small set of states for policy questions that involve overlapping sources. The labels can vary, but the intent should be clear:

  • Resolved: one authorized source clearly governs the question.
  • Conditionally resolved: the answer applies only within stated scope, dates, or conditions.
  • Needs owner review: two relevant sources remain, but authority or precedence is unclear.
  • Insufficient information: a required fact is missing, such as an approval date or employee scope.
  • Escalated: the question has been assigned to a named HR policy owner for a decision.

This prevents a retrieval result from masquerading as a final policy interpretation. It also gives HR a consistent way to review patterns. If the same attendance exception question is repeatedly escalated, the underlying policy language or approval process may need clarification.

Record the human decision without exposing private information

When a policy owner resolves the issue, the record should preserve the decision’s business meaning:

  • Which sources were reviewed.
  • Which scope and effective-date facts mattered.
  • Who owned the decision.
  • What interpretation was approved.
  • Any limits, expiration date, or required follow-up.
  • Whether the result applies only to this case or should clarify a broader policy question.

The record should contain only the employee information needed for the decision and future audit. It should not turn a policy review into a broad disclosure of private employee details.

A later assistant response can then distinguish between the original documents and the owner’s recorded interpretation. That distinction is valuable. The documents are evidence; the authorized decision explains how they apply in this case.

Why this is a governance issue

The NIST AI Risk Management Framework is voluntary guidance intended to help organizations manage risks associated with AI and incorporate trustworthiness considerations into the design, development, use, and evaluation of AI systems. Its relevance here is practical: an HR team should be able to identify uncertainty, assign responsibility, and evaluate how an AI-assisted process behaves when the source material does not produce one obvious answer.

NIST’s framework is not an HR policy and does not decide which workplace rule controls. It is useful governance context, not a substitute for the organization’s own policy ownership, legal review, or employee-relations judgment.

For HR leaders, the operating principle is simple: an assistant may organize evidence, but an authorized owner must decide unresolved policy authority. That principle supports accountability without requiring the assistant to make employment decisions.

What a responsible operator should do next

Start with one recurring overlap, such as schedule exceptions and attendance. Ask five questions:

  • Can the assistant show every approved source it considered relevant?
  • Can it distinguish document relevance from authority to override or modify a rule?
  • Does it surface scope and effective dates before producing a conclusion?
  • Does it have a defined pause and routing behavior when authority is unclear?
  • Is there a named HR owner who can record the decision and its limits?

Test the process with deliberately ambiguous examples. The goal is not to make the assistant answer every question. The goal is to make uncertainty visible early enough that an employee does not mistake an incomplete interpretation for permission or a compliance instruction.

The takeaway

When two approved policies appear to apply, the right question is not “Which answer sounds most likely?” It is “Who has authority to determine which rule governs this case?”

An AI assistant should identify both sources, preserve the uncertainty, state the relevant scope and dates, and route the unresolved question to the designated HR policy owner. That keeps retrieval useful, decisions accountable, and employees from receiving a silent interpretation that no human authorized.

When Two Approved Policies Apply, AI Should Not Decide in Silence - sharing image
When Two Approved Policies Apply, AI Should Not Decide in Silence - inline comparison
When Two Approved Policies Apply, AI Should Not Decide in Silence - inline explainer
Avery, your AI guide

Your role. Your questions.

Need HR guidance?
Ask Avery.

Talk through an idea, ask about the tools you already use, or find out what a first project could look like.

Avery is a fictional campaign character and AI guide. Our team handles demo requests.