Overview

You ask an AI to draft an internal email about next month's team lunch. It comes back in two minutes. You don't verify it. You don't run it through a checklist. You send it. That's right. You shouldn't.

You ask an AI to help evaluate a vendor for a $2 million contract. It takes an hour. You verify it thoroughly. You have another team member review it. You check the risk assessment three times. You document what you approved. That's also right.

The difference isn't the AI. It's the materiality of the decision. Some AI output hardly matters if it's wrong. Other AI output could cost you significant money, expose you to risk, or create liability if it's wrong. The amount of verification you apply should match the actual risk of the decision.

This is materiality-based verification. And it's how professional operations teams work with AI.

Materiality: When Does AI Output Actually Matter?

Materiality is an auditing concept. Something is material if getting it wrong would change a decision or cause meaningful harm. It's not about perfect accuracy. It's about accuracy that matters.

In operations, materiality has four dimensions:

Financial Materiality

How much money is involved? A $500 error in a $100,000 vendor contract is material. A $500 error in next year's budget estimate is immaterial (that's just planning variance).

For your organization, define financial thresholds. These examples are for a mid-size operations group:

• Immaterial: Under $10,000 (minor spend, decision unlikely to change)
  - Low materiality: $10,000-$50,000 (noticeable but not decision-changing)
  - Medium materiality: $50,000-$250,000 (significant, requires careful review)
  - High materiality: $250,000+ (major decision, requires extensive verification)

Your thresholds depend on your organization's size and risk appetite. For a $10M ops budget, $100K is very material. For a $100M ops budget, $100K is low-medium materiality.

Compliance Materiality

Does getting this wrong violate regulations or expose you to liability? A minor compliance gap might be immaterial. A violation that could trigger an audit is material.

For your operations, identify what's material:

• Immaterial: Best practice recommendations that don't affect compliance (nice-to-haves)
  - Low materiality: Minor process improvements that might improve compliance posture slightly
  - Medium materiality: Recommendations affecting compliance but won't trigger audits if missed (important but not critical)
  - High materiality: Actions directly affecting whether you're in compliance or at audit risk (critical)

Examples: A formatting issue in an SOP is immaterial. A missing audit trail in a process touching financial records is high materiality.

Operational Materiality

How many people does this affect? How critical is the process? A mistake in a procedure three people use is less material than a mistake in a process 200 people depend on.

• Immaterial: Affects 1-3 people; low-criticality process (team can work around it)
  - Low materiality: Affects 3-20 people; standard criticality
  - Medium materiality: Affects 20-100 people OR high-criticality process (people can't work without it)
  - High materiality: Affects 100+ people OR critical process (entire operation depends on it)

Examples: A procedure error in a 5-person team is low materiality. A procedure error in order fulfillment (100+ people rely on it) is high materiality.

Reputational Materiality

Would getting this wrong damage your credibility or customer relationships? A mistake in a vendor evaluation affects your judgment. A mistake in a customer-facing communication affects your reputation.

• Immaterial: Internal process; mistakes don't affect customer trust or external perception
  - Low materiality: Internal communication; mistakes affect team perception (team loses confidence in you)
  - Medium materiality: Affects customer perception or affects trust with leadership (customer notices, or CFO loses confidence)
  - High materiality: Public-facing or affects customer trust significantly (customer complaint, public issue, credibility damage)

Examples: An internal process doc error is low materiality. A public customer announcement with errors is high materiality.

Four Verification Tiers: Match Verification to Risk

Once you know materiality, apply the right verification level. Four tiers cover almost everything:

Tier 1: Minimal Verification (Low Risk, Immaterial)

When to use: AI output on low-materiality decisions. Internal emails. Brainstorming documents. First-draft guidance on non-critical topics. Suggestions you'll significantly revise anyway.

What you do:

• Skim the output (2 minutes)
  - Check for obvious errors (bad grammar, nonsensical suggestions)
  - Make sure the tone is appropriate for the audience
  - Use it as-is or with minor edits
  - No documentation required

Example: AI drafts an internal message about office supply ordering changes. You check that it's clear, sounds professional, and doesn't say anything wrong. You send it. This is Tier 1.

Your verification checklist (tier 1) - 2 minutes:

>
- [ ] No obvious grammar errors
- [ ] Tone is appropriate
- [ ] Key point is clear
- [ ] No harmful or misleading statements

Tier 2: Standard Verification (Medium Risk, Low-Medium Materiality)

When to use: Routine operational work. Standard SOPs. Regular vendor analysis. Process documentation. Most operational output you produce regularly.

What you do:

• Use your standard verification checklist (5-10 minutes)
  - Spot-check key claims or facts (verify 2-3 important statements)
  - Verify it matches your standards and templates
  - Approve or request revisions
  - Brief documentation: "Reviewed and verified" with date

Example: AI drafts a standard vendor evaluation for a $50K vendor. You check completeness (all criteria covered?), spot-check a couple of cost figures, confirm compliance status looks right, ensure recommendation makes sense. Takes 10 minutes. You approve it.

Your verification checklist (tier 2) - 10 minutes:

>
COMPLETENESS
- [ ] All required sections present
- [ ] All evaluation criteria covered

ACCURACY
- [ ] Spot-check 3 factual claims
- [ ] Math checks out
- [ ] References are accurate

CONSISTENCY
- [ ] Follows our format
- [ ] Uses our terminology
- [ ] Recommendation aligns with findings

USABILITY
- [ ] Clear recommendation
- [ ] Actionable next steps

Tier 3: Enhanced Verification (High Risk, Medium-High Materiality)

When to use: High-materiality output. Major vendor decisions ($250K+). Significant process changes. Compliance-critical documentation. Large financial decisions ($100K+).

What you do:

• Use comprehensive verification checklist (20-30 minutes)
  - Verify all major claims, not just spot-checks
  - Have a colleague review key findings
  - Document what you verified and approved
  - Request revision if issues found

Example: AI analyzes a major vendor switch for your supply chain ($300K annually). You fully verify costs (check 3 price quotes), compliance (confirm certifications are real), risks (identify all supply chain risks). A colleague reviews the risk assessment. You document: "Reviewed by [you], [date]. Verified costs, compliance, risks. Approved for decision." Only then do you take it to leadership.

Your verification checklist (tier 3) - 30 minutes:

>
ACCURACY
- [ ] Verify all major claims (not spot-check)
- [ ] Check all numbers and math
- [ ] Verify all regulatory references
- [ ] Have colleague verify key finding

COMPLETENESS
- [ ] Missing anything?
- [ ] All edge cases considered?

RISK ASSESSMENT
- [ ] Is risk assessment thorough?
- [ ] Are mitigations realistic?
- [ ] Have we missed any risks?

DECISION SUPPORT
- [ ] Is recommendation justified?
- [ ] Are alternatives considered?
- [ ] Is implementation plan realistic?

DOCUMENTATION
- [ ] Verification documented
- [ ] Date and scope recorded
- [ ] Who verified listed

Tier 4: Expert Review (Critical Risk, High Materiality with Compliance/Reputational Risk)

When to use: Critical decisions. Anything involving legal or regulatory risk. Situations where being wrong has severe consequences. Major strategic decisions. Compliance-critical procedures.

What you do:

• Run full verification checklist (30+ minutes)
  - Have SME (subject matter expert) or compliance expert review
  - Consider getting external validation if needed
  - Document thoroughly for audit trail
  - May reject AI assistance entirely on critical items and do it yourself

Example: AI drafts a new compliance procedure for healthcare operations involving HIPAA. You fully verify. Your HIPAA compliance officer reviews it. You decide: "AI can draft this, but my compliance officer will write the final version." You document the full verification process.

Your verification checklist (tier 4) - 45+ minutes:

>
All tier 3 checks, plus:
- [ ] SME review completed
- [ ] Legal/compliance clearance if needed
- [ ] External validation if high-risk (vendor check, regulatory agency confirmation?)
- [ ] Alternative approaches considered
- [ ] Failure modes analyzed
- [ ] Sign-off from decision maker
- [ ] Comprehensive audit trail
- [ ] Documentation of who reviewed and when

When NOT to use AI at all

Tier 4 sometimes means "don't use AI for this at all." Not everything should be AI-assisted. Some decisions are too important, too risky, or too nuanced. If using AI on something critical would require verification so extensive that you might as well have done it yourself, the answer might be: don't use AI. That's a valid decision. Part of AI proficiency is knowing when not to use it.

Building Your Materiality Framework

Create a clear decision framework for your operations. This should be a document your team can reference:

Step 1: Identify Material Decision Categories

What types of decisions do you make regularly? For each, assess materiality:

>
DECISION TYPE: Vendor evaluation
FINANCIAL MATERIALITY: Depends on contract value
- Under $50K: Low
- $50K-$250K: Medium
- Over $250K: High
COMPLIANCE MATERIALITY: Medium (vendor compliance affects our compliance)
OPERATIONAL MATERIALITY: Medium (affects multiple teams)
REPUTATION MATERIALITY: Low-Medium (internal decision)
OVERALL TIER: Mostly Tier 2, escalate to Tier 3 for contracts over $250K

>
DECISION TYPE: Process SOP writing
FINANCIAL MATERIALITY: Low (doesn't directly cost money if wrong)
COMPLIANCE MATERIALITY: HIGH (SOP compliance affects our compliance)
OPERATIONAL MATERIALITY: Medium-High (team depends on SOP; wrong SOP breaks process)
REPUTATION MATERIALITY: Medium (SOP quality reflects on operations team)
OVERALL TIER: Tier 2 for standard SOPs, Tier 3 if touching compliance-critical process, Tier 4 if safety-critical

>
DECISION TYPE: Cost reduction initiative analysis
FINANCIAL MATERIALITY: Depends on savings
- Under $10K: Low
- $10K-$100K: Medium
- Over $100K: High
COMPLIANCE MATERIALITY: Medium-High (must not compromise safety or compliance)
OPERATIONAL MATERIALITY: Depends on scope (affects how many people?)
REPUTATION MATERIALITY: Low
OVERALL TIER: Tier 2 for small initiatives, Tier 3 for significant cost reductions, Tier 4 if affecting safety/compliance

Step 2: Assign Tiers to Regular Tasks

For work you do repeatedly, pre-assign the tier:

>
Routine vendor evaluations (under $100K): Tier 2
Significant vendor decisions (over $100K): Tier 3
Internal process documentation: Tier 2
Compliance-related procedures: Tier 3
Customer-facing communications: Tier 2-3
Strategic planning documents: Tier 3
Employee communications: Tier 1-2
Budget analysis: Tier 2-3 (depends on amount)
Safety procedure updates: Tier 4
Vendor contract negotiations: Tier 3-4 (depends on size)

Step 3: Train Your Team

Share your materiality framework with your team. "Here's how we decide what level of verification to apply. Use this when you're reviewing AI output."

This prevents two problems: (1) Over-verification of low-risk items (wasting time), and (2) Under-verification of high-risk items (creating risk).

Decision Matrix: When to Skip AI Entirely

Sometimes the right answer is: don't use AI for this. Here's when:

Skip AI when:

• The verification would take as long as doing it yourself. If you'd spend 3 hours verifying an AI output, you might as well spend 2 hours doing it yourself without AI. This is Tier 4 with extensive external review.
• The decision requires judgment that's inherently human. Evaluating someone's performance. Making strategic choices about company direction. Some decisions need your judgment, not AI assistance. AI can provide data, but the decision is yours.
• The context is so specific to your company that AI can't understand it. If the AI needs to understand years of institutional history, internal politics, or nuanced company culture, it probably can't get it right. Don't use AI for highly contextual decisions.
• The risk of being wrong is catastrophic and verification is uncertain. If getting it wrong could cause serious harm and you can't reliably verify AI work, skip AI. Example: Medical compliance procedures where AI hallucination could cause HIPAA violation.
• You don't trust the AI's domain knowledge. If the AI regularly makes mistakes in a specific domain, don't use it for high-stakes decisions in that domain. Use it for other things where you're confident.
• The decision will be scrutinized later. If you'll need to explain this decision to auditors or leadership and they'll question whether AI assistance was appropriate, be cautious. Consider doing it yourself or having heavy expert involvement.
• Legal or regulatory requirement says a human must do this. Some decisions can't be delegated to AI. Know your regulatory environment. If compliance requires "human review and approval," you can't skip that for any decision, no matter how low-risk.

This is important: proficiency with AI includes knowing when NOT to use it. Not every task benefits from AI assistance. Sometimes the human approach is faster, safer, or more appropriate.

Try This Now: Build a Materiality Framework for Your Operations

Take 45 minutes. Create a materiality framework for your organization.

Step 1: List your regular decision types. What do you ask AI to help with regularly? Vendor analysis? Process documentation? Reports? Compliance work? Budget analysis? List 8-10 categories.

Step 2: For each decision type, assess materiality. Use the four dimensions: Financial, Compliance, Operational, Reputational. Rate each as Low, Medium, or High.

Step 3: Assign verification tier. Based on materiality, assign Tier 1-4. Or decide: skip AI for this entirely.

Step 4: Document it. Create a simple table your team can reference.

Step 5: Share and refine. Get feedback from your team. Does the framework make sense to them? Does it match how you actually work? Adjust based on input.

Example framework (healthcare operations):

>
DECISION | FINANCIAL | COMPLIANCE | OPERATIONAL | REP | TIER | NOTES
---|---|---|---|---|---|---
Vendor compliance verification | Low-Med | HIGH | Med | Med | Tier 3 | Compliance officer review required
Routine SOP updates | Low | Med | Med | Low | Tier 2 | Use standard checklist
New vendor eval (under $50K) | Low | Low | Low-Med | Low | Tier 2 | Standard verification
Major vendor (over $250K) | HIGH | Med | High | Med | Tier 3-4 | Enhanced verification + approval
Process improvement suggestions | Low | Low | Low | Low | Tier 1-2 | Tier 1 for brainstorm, Tier 2 for impl
Safety procedure updates | Low-Med | HIGH | HIGH | High | Tier 4 | Safety committee review required
Staff communication | Low | Low | Low | Low | Tier 1 | Skim for tone
Patient-facing documentation | Low | HIGH | High | HIGH | Tier 4 | Compliance + communications review
Budget forecast | Low-Med | Low | Low | Low | Tier 2 | Spot-check math
Cost reduction analysis | Med-High | Med-High | Med | Low | Tier 2-3 | Tier 3 if affecting safety/compliance

What to Do Monday Morning

• Create your materiality categories. What matters most in your operations? Financial impact? Compliance? Operational criticality? Reputation? Define your thresholds.
• Assess your regular AI tasks. When you ask AI for help, what's the materiality? Build a quick reference table.
• Assign verification tiers. For each task, decide: Tier 1, 2, 3, or 4? Or skip AI entirely?
• Create checklists for Tier 2 and Tier 3. You'll use these repeatedly. Make them specific to your work type.
• Share the framework with your team. Make this a team resource. "Here's how we decide how much to verify AI output."
• Start applying it next week. Next time you get AI output, check the materiality framework. Use the right tier. Notice how much time you save and how your confidence in the work increases.

Key Takeaways

• Not all AI output needs the same verification. Match verification to materiality. Low-risk decisions need quick review. High-risk decisions need thorough review. The effort should fit the consequence.
• Materiality has four dimensions: financial, compliance, operational, reputational. Rate each. That tells you how carefully to verify.
• Use four verification tiers: Minimal, Standard, Enhanced, Expert. Tier 1 for low-risk, low-materiality work (2 minutes, no documentation). Tier 4 for critical work (45+ minutes, full documentation).
• Create a materiality framework for your organization. Decision types, materiality assessment, assigned tier. Make it a team resource. Reference it every time you get AI output.
• Know when to skip AI entirely. If verification would take as long as doing it yourself, or if the risk is too high, don't use AI. That's proficiency, not laziness.
• Materiality frameworks are living documents. Review them quarterly. When something goes wrong and you wish you'd verified more carefully, that's a signal to increase materiality for that decision type.
• Speed is valuable, but accuracy matters more. Don't skip verification on high-materiality decisions just to move faster. That's not speed, that's risk.

Frequently Asked Questions

Q: How do I know if something is truly high materiality or if I'm just nervous about the decision?

A: Ask yourself: If the AI got this completely wrong, what would happen? Would you lose significant money? Would you violate compliance? Would your team be unable to work? Would customers lose trust? If the answer to any of these is yes and the amount is large, it's high materiality. Your nervousness is often correct, but back it up with impact analysis.

Q: Can I use Tier 1 verification for something I later realize should have been Tier 3?

A: Unfortunately, you can't take back a published error. That's why your initial assessment matters. If you're uncertain about materiality, go one tier higher. It takes 5 extra minutes and protects you. Better to over-verify once and refine your framework than to under-verify and create a problem.

Q: What if my CEO wants me to speed up decision-making by using AI without much verification?

A: This is a real tension. Speed is valuable, but so is accuracy. Use your materiality framework to distinguish high-speed/low-risk work from careful/high-risk work. AI can make you much faster on low-risk decisions. On high-risk decisions, explain: "We can go faster with AI, but we need to verify thoroughly. Here's the trade-off." Let leadership choose. But don't skip verification on something that should have Tier 3 or 4 just to move faster.

Q: Does materiality change over time? Should I revisit my framework?

A: Yes. If your organization grows, materiality thresholds change. If regulations change, compliance materiality changes. Review your framework annually or whenever your business changes significantly. When something goes wrong, that's a signal to increase materiality for that decision type.

Q: Is there a risk of bias in assessing materiality? Might I rate something as lower materiality because I like the AI's output?

A: Yes, that's a real risk. To counter it: (1) Assess materiality based on decision type, not the specific situation. (2) Do materiality assessment before you see the AI output. (3) Have someone else challenge your materiality rating if they think you're being too lenient. Building in a second opinion helps catch bias.