Overview

Lecture URL: https://skill.re/learn/manager/documenting-ai-use-in-team-processes.php

AI FOR MANAGERS CERTIFICATION

Responsible AI Oversight (Level 2) | Responsible AI Oversight

LECTURE: Documenting AI Use in Team Processes

Lesson 2.5.5 | Estimated Duration: ~22 minutes

Welcome to the AI for Managers certification program. I am your instructor, and today we are covering one of the most practical lessons in the Responsible AI Oversight module: Documenting AI Use in Team Processes.

This is Lesson 2.5.5 in Level 2, the Responsible AI Use track. Whether you are implementing AI tools across your team, ensuring compliance with regulations, or building institutional knowledge, the material in this session is designed to meet you where you are.

Today we focus on a sometimes-overlooked aspect of responsible AI use: documenting where and how AI is used. This serves multiple purposes: compliance, team clarity, risk management, and learning.

The goal is not to create excessive bureaucracy. The goal is to create the minimum documentation that serves important purposes.

Before we begin, I encourage you to think about which of your team's processes use AI and whether that is documented anywhere. By the end of this session, you will have frameworks for practical documentation.

Let us get started.

Lesson 2.5.5: Documenting AI Use in Team Processes

Purpose

When AI becomes embedded in team processes, it is no longer optional to document. Documentation serves compliance, clarity, accountability, and continuous improvement.

This lesson teaches you what to document and how to do it efficiently.

Why This Matters for Managers

A common mistake is not documenting AI use, then discovering later that:

• You cannot explain to a client how a deliverable was created
  - Regulatory auditors ask questions and you have no answers
  - Someone uses AI inappropriately and you have no trail to catch it
  - A new team member does not know which processes use AI
  - You want to improve a process but cannot remember how it works

The stakes include:

• Compliance (for regulated industries, documentation might be required)
  - Accountability (you need to answer for decisions and outputs)
  - Risk management (if something goes wrong, documentation helps investigation)
  - Onboarding (new team members need to understand how work is done)
  - Learning (you improve processes when you understand them)

Your job is to establish practical documentation standards.

What to Document

Level 1: Process Documentation

For each process that uses AI, document:

1. What is the process? (name, purpose, who does it)
2. Where is AI used? (which step)
3. What AI tool is used? (which tool, which features)
4. Why AI? (what problem does it solve, what value does it create)
5. Verification/review: (how is AI output checked before use)
6. Limitations: (what can go wrong, what is this AI not good at)

Format: One-page summary per process

Example: Email Campaign Process

"Process: Customer Email Campaign

Purpose: Send monthly promotional emails to customers

Who: Marketing team

Steps:

1. Research topic for month -> Human
2. Create email list -> Human
3. Draft email -> AI (using ChatGPT Pro, prompt engineering for tone/message fit)
4. Review and edit email -> Human (one person, typically Sarah)
5. A/B test variants -> AI (generating 3 subject line options)
6. Send -> Human (approval required before sending)

Why AI?

• Email drafting: Saves 30 minutes per email
  - Subject line testing: Generates options faster than human brainstorming
  - Preserves tone and brand voice through careful prompting

Verification:

• All drafts reviewed by human for accuracy and tone
  - Facts verified against marketing database
  - No customer personal data input to AI tool

Limitations:

• AI can be overconfident about statistics. Always verify facts.
  - AI-generated emails sometimes miss cultural nuance. Sarah catches this in review.
  - A/B test options generated by AI are suggestions only. Human final decision.

Compliance notes:

• No GDPR-sensitive data input to tool
  - Email source code stays within our systems
  - AI is helper, not decision maker"

Level 2: Output Documentation

For outputs created with AI assistance, document:

1. What is the output? (email, report, code, etc.)
2. What AI was used? (if any)
3. What verification happened? (review checklist, who reviewed)
4. Is it final or draft? (if draft, what is still needed)

Format: Brief comment or annotation on the output itself

Example documentation in email:

"This email was drafted with AI and reviewed by Sarah for tone and accuracy. It reflects our standard promotional language and has been verified against current product specifications. Ready to send."

Example documentation in code review:

"This code was reviewed with AI tool support (GitHub Copilot). AI identified three potential issues (flagged in comments). All flagged items were manually reviewed and confirmed/adjusted by developer. This is final."

Level 3: Decision Documentation

For important decisions involving AI, document:

1. What was the decision? (use/not use AI, choose which tool, etc.)
2. What was the reasoning? (why this decision)
3. What was considered? (alternatives, trade-offs)
4. Who made the decision? (person/team accountable)

Format: Brief memo or decision log entry

Example: Tool Selection Decision

Reasoning: Tool best fit our security requirements (enterprise SOC 2, data not used for training), budget constraints ($X/month), and integration with our GitHub workflow

Considered: Option A (lower cost but privacy concerns), Option B (better features but significantly higher cost)

Outcome sought: Reduce code review time by 20-30% while maintaining quality

Decision made by: Engineering director + team lead

Creating a Documentation System

Step 1: Identify What Needs Documentation

Audit your team processes:

• Which processes use AI?
  - Which are important enough to document? (usually those that touch customer deliverables, compliance-sensitive work, or significant time investment)
  - Which decisions about AI use should be documented?

Rule of thumb: If you would want to explain it to a regulatory auditor or a new team member, document it.

Step 2: Create Templates

Do not force lengthy documentation. Use templates that are quick to fill out.

Template 1: Process Use Summary (one-page max)

Template 2: Output Documentation

"AI used: Yes / No

Status: Draft / Final"

Template 3: Decision Log

Step 3: Choose Where to Document

You need documentation in places where team members will actually see it and use it:

Option 1: Process Documentation Wiki

• Centralized location for all process documentation
  - Easy to search and update
  - Can be version-controlled
  - Tools: Confluence, Notion, GitHub wiki, SharePoint

Option 2: Embedded in Process Workflows

• Documentation lives where the process lives
  - In Slack channels, issue tracking systems, project management tools
  - Easier to keep up to date because it is where people work

Option 3: Decision Log

• Separate from process documentation
  - Captures decisions over time
  - Useful for retrospectives and learning
  - Can be simple: spreadsheet, Notion database, etc.

Best practice: Use a combination. Process documentation in wiki, annotations/comments in actual workflows, decisions in a log.

Step 4: Make Documentation a Habit

Documentation is only useful if it stays current. Make it a routine:

When launching a new process with AI:

• Draft documentation at the same time as process design
  - Share with team: "Here is the process AND how we document it"
  - Update as you learn what works

When you change a process:

• Update documentation immediately
  - Note what changed and why
  - Share the update with the team

When you hire new people:

• Have them read documentation
  - Have them suggest clarifications (if they do not understand, it needs fixing)
  - Update based on feedback

Quarterly (or semi-annually):

• Review documentation for accuracy
  - Update process descriptions if processes have evolved
  - Archive old documentation

Handling Special Cases

Case 1: Client Work with AI

When you deliver client work created with AI assistance, be transparent:

• In contract/agreement: If AI was used, disclose it
  - In deliverable: Note what AI was used, how it was verified
  - In communication: Explain your quality assurance process

Why this matters:

• Client trust depends on transparency
  - Some clients might have concerns about AI and need assurance
  - Regulatory compliance in some industries requires disclosure

Case 2: Regulated Industry Work

For healthcare, finance, legal, or other regulated industries:

Check regulations:

• Some regulations require documenting all decision-making processes
  - Some have specific requirements for AI use disclosure
  - Some have audit requirements

Document more thoroughly:

• What data was input to AI (because data handling matters)
  - Who reviewed and approved output (chain of verification)
  - How you ensured output meets regulatory standards
  - What governance process approved AI use for this work

Example: HIPAA-regulated healthcare environment

"Patient summary created with AI assistance

Verification process: Clinician reviewed all clinical details against original records

This summary used AI for organization and highlighting of key points. Clinical judgments and diagnoses are clinician-verified, not AI-generated."

Case 3: Internal Process with Low Risk

For internal processes where risk is minimal:

Document lightly:

• Process name and purpose
  - Which tool is used
  - Who maintains it
  - That is often enough

Example: Internal report writing

"Sales report template

Uses: AI text summarizer to convert detailed notes into executive summary

How it works: Paste notes into ChatGPT Plus, use standard prompt, review output for accuracy, include in report

Time saved: ~20 minutes per report

Owner: Sales manager"

Case 4: Evolving Processes

As you experiment with AI, processes evolve. Your documentation should reflect that:

Early stage:

• "This is in pilot phase"
  - "Owner testing, not yet team-wide"

Established:

• "All team members use this approach"
  - "Documented and verified"

Sunset (moving away from):

• "This process is being replaced"
  - "No longer recommended for new work"

Documentation as Learning Tool

Beyond compliance and clarity, documentation helps you improve:

After 3 months of using a process with AI:

• Review the documentation
  - Compare it to reality: "Are we really doing this?"
  - Assess: "Is this working? Could we do it better?"
  - Update: "Here is how we are refining the process based on learning"

When someone suggests improvements:

• Document the suggestion and reasoning
  - Test it
  - If it works, update the process documentation

Documentation creates a record of evolution, which is valuable for continuous improvement.

Common Documentation Mistakes

Mistake 1: Over-Documenting

"We have 50-page documentation for every process."

Why it fails: Too much documentation becomes stale. People stop reading it. It stops being useful.

Better: One-page summary per process. Links to more detail if needed. Keep it light.

Mistake 2: Under-Documenting

"We don't document anything. People just know how processes work."

Why it fails: New people have to learn by trial and error. Knowledge is fragile. If someone leaves, processes break.

Better: Document enough that a new team member can follow the process with minimal training.

Mistake 3: Documenting Without Updating

"We documented the process once and never touched it again."

Why it fails: Processes evolve. Documentation becomes out of date. People follow old processes or ignore documentation because it is wrong.

Better: Document once, then update regularly. Build updating into your routine.

Mistake 4: Making Documentation Too Technical

"The documentation assumes technical knowledge that non-technical people don't have."

Why it fails: Team members cannot follow it. It is useless.

Better: Write for your actual audience. Use plain language. Define technical terms.

Mistake 5: Not Involving the Team

"I documented the processes without asking the team."

Why it fails: Documentation might miss nuances people know. Team does not buy in.

Better: Draft documentation, then have team review and suggest changes. "Is this accurate? What am I missing?"

ANTI-PATTERNS

Anti-Pattern 1: Treating Documentation as Compliance Checkbox

"We documented AI use because regulations require it. Nobody actually reads it."

Why it fails: If documentation is not useful to the team, it is not maintained. It becomes stale.

Better: Documentation serves the team first. It helps them work better. Compliance is a bonus.

Anti-Pattern 2: Separate from Actual Work

"Documentation is in a wiki. Actual work happens in Slack."

Why it fails: People do not consult documentation if it is not where they work.

Better: Documentation lives where work happens, or links to where work happens.

Anti-Pattern 3: Blaming Documentation When Things Go Wrong

"You should have read the documentation" (said when someone made a mistake)

Why it fails: This creates defensiveness, not learning. People feel blamed.

Better: Use mistakes as opportunities to improve documentation. "I see the documentation was unclear. Let's fix it."

Anti-Pattern 4: No Accountability for Updates

"Documentation is everyone's job, so nobody updates it."

Why it fails: No ownership means it does not get updated.

Better: Assign ownership. "Sarah owns the documentation for this process. Sarah, let's update it quarterly together."

Anti-Pattern 5: Documentation Becomes Compliance-Only

"We only document to satisfy auditors."

Why it fails: Documentation for compliance is usually dry and unhelpful for day-to-day work.

Better: Documentation serves the team. Compliance is a benefit.

PRACTICE PROMPTS

1. Process Audit: List three processes your team uses that involve AI. For each, what level of documentation do you have now? What should you have?
2. Template Creation: Using the templates provided, draft documentation for one process your team uses.
3. Documentation System: Design how you would store and maintain documentation for your team. Where would it live? Who would maintain it? How often would it update?
4. Client Communication: If you were delivering AI-assisted work to a client, what would you want to tell them about how it was created and verified?
5. Improvement Loop: Document a process your team uses. In 3 months, revisit the documentation. What has changed? What would you improve?

KEY TAKEAWAYS

1. Document processes that use AI, including what AI is used, why, and how output is verified.
2. Keep documentation practical and light. One-page summaries work better than lengthy docs.
3. Place documentation where people actually work. Separate wikis are useful supplements, not replacements.
4. Update documentation regularly. Build updating into your routine.
5. Use documentation as a learning tool. Update processes based on what you learn, then update documentation.

GLOSSARY

Process Documentation: A written description of how a process works, including steps, tools, and people involved.

Verification Trail: A record showing how an output was created, reviewed, and approved.

Compliance Documentation: Documentation created to meet regulatory or organizational requirements.

Decision Log: A record of important decisions, reasoning, and outcomes. Useful for learning and continuity.

Owner (or Process Owner): The person responsible for maintaining and improving a documented process.

[SYNTHESIS AND APPLICATION]

Let us step back and look at the bigger picture of what we have covered in this session on Documenting AI Use in Team Processes.

Documentation feels like overhead. But it serves multiple purposes: compliance, clarity, continuity, and learning.

Here is what I want you to take away from this session:

First, simplicity. Keep documentation light. One-page summaries, not lengthy manuals.

Second, utility. Documentation works when it serves the team. Keep it where people work.

Third, evolution. Documentation should evolve as processes improve. Use it as a learning tool.

[REFLECTION EXERCISE]

Before we close, I would like you to spend two minutes on this reflection:

Think about a process your team uses. If a new team member joined tomorrow, could they understand how it works from documentation? If not, what would need to be documented?

Write down your answer. That reflection guides your documentation priorities.

[CLOSING REMARKS]

In our next lesson, we will explore recognizing AI errors and building habits to catch them. Documentation supports this: when you document how processes work, you can also document what to watch for and how to catch errors.

This has been Lesson 2.5.5: Documenting AI Use in Team Processes, part of the Responsible AI Oversight module in Level 2: Responsible AI Use of the AI for Managers certification.

Remember: documentation is not bureaucracy. It is a tool for clarity, learning, and accountability.

Thank you for your time, your attention, and your commitment to documenting and continuously improving how your team uses AI.

END OF TRANSCRIPT

AI for Managers Certification Program

Level 2: Responsible AI Use | Responsible AI Oversight | Lesson 2.5.5

A SkillsClinic initiative by No Worker Left Behind and The Work Company.

Duration: ~22 minutes | Word Count: ~3521