Building AI-Assisted Workflows — Foundations Lesson 11

Scenario

You've learned prompting, verification, ethics, security, and format specification. But you still approach each task differently. One day, you realize: I could create a repeatable workflow for routine tasks. Instead of figuring out how to use AI each time, you could have a process. Instead of hoping you remember to verify output, verification could be built in. Instead of using different tools and prompts, you could have a consistent approach. This is the difference between using AI occasionally and using it professionally.

---

Learning Objectives

By the end of this lesson, you will be able to:

1. Map out a complete workflow — from task start to final output

2. Identify AI's role in each step (automate, support, skip)

3. Build verification into process — so checking happens automatically

4. Create templates and prompts you can reuse

5. Document workflows so others can use them (and you can repeat them)

---

Think Before You Design

Before you build a workflow, ask:

What's the outcome I need?

Which steps are routine (same every time)?

Where could AI add the most value?

What requires human judgment?

Where do errors typically happen?

How would I know if this worked?

Could someone else follow this process?

---

Workflow Components

Every AI-assisted workflow has these components:

1. Input (What You Start With)

What it is: The information or raw material that starts the process

Example: Customer complaint email, meeting notes, project requirements

Questions to ask:

What format is input in?

How much variation is there?

What information is always present? What's sometimes missing?

Should you standardize input first?

---

2. AI Processing (Where AI Helps)

What it is: The specific AI task that processes input

Example: Summarizing email, categorizing complaint, extracting data

Questions to ask:

What exactly should AI do here?

Is AI automating or supporting?

What prompt would work?

What's the output format?

How will I verify it?

---

3. Verification/Review (Catching Problems)

What it is: The step where you check AI output

Example: Spot-checking accuracy, reviewing tone, confirming completeness

Questions to ask:

What could go wrong here?

What's the risk level?

How much review is needed?

What are you checking for?

Could this be automated (spell-check, grammar)?

---

4. Refinement (If Needed)

What it is: Iterating or improving AI output

Example: Adjusting tone, adding missing information, correcting errors

Questions to ask:

How often does output need refinement?

What refinements are common?

Could you prompt better to avoid refinement?

Could refinement be templated (known fixes)?

---

5. Integration/Output (Where Result Goes)

What it is: Using the final output (sending, publishing, storing, etc.)

Example: Sending email, publishing report, updating system

Questions to ask:

Where does output go?

Is there a final approval step?

Are there compliance/quality gates?

How is this documented or tracked?

---

Workflow Template: Basic to Advanced

Simple Workflow (3 steps)

`

INPUT → AI PROCESS → OUTPUT

`

Example: "Quick Email Draft"

1. Input: Customer request email

2. AI process: Draft response email

3. Output: Send to customer

(Minimal verification — low-risk scenario)

---

Standard Workflow (5 steps)

`

INPUT → AI PROCESS → VERIFY → REFINE (if needed) → OUTPUT

`

Example: "Weekly Report"

1. Input: Team member updates

2. AI process: Compile into report format

3. Verify: Scan report for completeness and accuracy

4. Refine: Adjust tone or add missing sections

5. Output: Distribute to leadership

---

Complex Workflow (7+ steps)

`

INPUT → PREP → AI PROCESS 1 → VERIFY 1 → AI PROCESS 2 → VERIFY 2 → REFINE → OUTPUT

`

Example: "Customer Onboarding"

1. Input: New customer signup

2. Prep: Gather customer data (name, company, needs)

3. AI process 1: Draft welcome email

4. Verify 1: Check personalization and tone

5. AI process 2: Generate resource guide

6. Verify 2: Check for relevance and accuracy

7. Refine: Adjust both documents for coherence

8. Output: Send email, provide guide

---

Building Your First Workflow

Step 1: Choose a Routine Task

Pick something you do regularly that takes 15-30 minutes.

Good candidates:

Weekly/monthly reporting

Email responses to common requests

Data summarization

Content adaptation

Meeting notes

Status updates

Avoid: One-time projects, highly creative work, decisions requiring deep judgment

---

Step 2: Map the Current Process

Write out how you do it now, step by step.

Example (Current Weekly Report Process):

1. Collect updates from 5 team members

2. Read through each update

3. Organize by theme

4. Write summary

5. Add metrics section

6. Proofread

7. Send to leadership

---

Step 3: Identify Where AI Helps

Go through your process and mark where AI could speed things up.

Example:

1. Collect updates from 5 team members → No, you need to do this

2. Read through each update → Possibly, AI could summarize

3. Organize by theme → Yes, AI could categorize

4. Write summary → Yes, AI could draft

5. Add metrics section → Maybe, AI could format data

6. Proofread → Partially, spell-check/grammar check

7. Send to leadership → No, you need to do this

---

Step 4: Design AI Steps

For each place AI helps, write a specific AI task.

Example:

Task 1 (Summarize): "Create a 1-sentence summary of each team member's update"

Input: Raw update text

AI prompt: "Summarize this update in 1 sentence focusing on accomplishments"

Output format: Bullet points

Verification: Read each summary; does it capture the key point?

Task 2 (Categorize & Draft): "Organize summaries by theme and draft report"

Input: 5 one-sentence summaries

AI prompt: [Template prompt for weekly report]

Output format: 2-3 paragraphs per theme

Verification: Does organization make sense? Is tone professional?

---

Step 5: Add Verification

For each AI step, define verification.

Risk-based approach:

Low-risk outputs: Quick scan (2 minutes)

Medium-risk outputs: Detailed review (5 minutes)

High-risk outputs: Expert review or testing (15+ minutes)

Example verification:

Summarization verification:

Quick scan (1 minute per summary)

Ask: Does this capture the key accomplishment?

Report verification:

Read full report (5 minutes)

Check: Organization is clear? Tone is professional? No hallucinations?

Quick source check: Does it accurately reflect what team said?

---

Step 6: Document the Workflow

Write it down so you can repeat it and teach others.

Format:

`

WORKFLOW: [Name]

PURPOSE: [What this workflow does]

TIME COMMITMENT: [How long it takes]

RISK LEVEL: [High/Medium/Low - based on impact if wrong]

INPUT:

[What starts this process]

STEPS:

Step 1: [What you do]

Step 2: [AI task]

Prompt: [Specific prompt or template]

Input: [What AI receives]

Output format: [Bullets/paragraphs/table/etc]

Verification: [What you check for]

Common issues: [What usually needs fixing]

Step 3: [Your verification/review]

What to check: [Specific criteria]

Time: [How long this takes]

When to ask for help: [Red flags]

Step 4: [Your refinement (if needed)]

What often changes: [Common tweaks]

How to refine: [Steps to improve output]

OUTPUT:

[Where this goes and how it's used]

REUSABLE ELEMENTS:

Template prompt: [For AI use]

Verification checklist: [What to check]

Common variations: [Adjustments for different scenarios]

`

---

Reusable Elements: Templates & Prompts

As you run workflows, save what works.

Template Prompts

Create prompts you use repeatedly. Save them for next time.

Example (Weekly Report):

`

You are a business analyst writing a weekly team report.

Context: Team of [number] members across [departments].

I receive individual updates from each team member.

Task: Organize these updates into a professional weekly report.

Format:

Introduction (1 paragraph)

[Department 1] accomplishments (1-2 paragraphs)

[Department 2] accomplishments (1-2 paragraphs)

Metrics (bullets)

Key risks/blockers (bullets)

Next week outlook (1 paragraph)

Tone: Professional, optimistic, factual

Word count: ~500 words

Audience: C-level executives

Here are the team updates:

[Updates pasted below]

`

Save this. Next week, just update the updates and paste.

---

Verification Checklist

Create a checklist of what you verify for each output type.

Example (Email Response Checklist):

[ ] Tone matches customer's tone (warm if they're warm, professional if formal)

[ ] Response addresses all questions asked

[ ] No jargon or unexplained terminology

[ ] Call to action is clear

[ ] Personalization (includes their name, references their situation)

[ ] Proofread for typos/grammar

[ ] Verification: Fact-checked any claims

Use this every time. Verification becomes automatic.

---

Common Workflow Patterns

Pattern 1: Summary → Distribute

Used for: Synthesizing information for others

Flow:

Collect raw information

AI summarizes

You verify

You distribute

Example: Meeting notes → AI summary → You verify → Send to team

Tools: Email, documents, Slack

---

Pattern 2: Template → Personalize → Send

Used for: Routine communications

Flow:

AI generates from template

You personalize

You verify tone

You send

Example: Customer onboarding → Email template → Personalize → Send

Key: Template varies little; personalization is what matters

---

Pattern 3: Intake → Categorize → Route

Used for: Processing inbound requests

Flow:

Receive request

AI categorizes/analyzes

You route to right person/process

Routing system continues

Example: Customer support ticket → AI priority flag → Route to team → Team handles

Key: AI helps with triage; humans handle resolution

---

Pattern 4: Data → Analysis → Presentation

Used for: Creating reports and insights

Flow:

Gather data

AI processes/analyzes

You verify analysis

You present findings

Example: Weekly metrics → AI summary → You verify → Present to stakeholders

Key: Verification is critical (analysis can be wrong)

---

Mission: Design Your Workflow

The Task

1. Choose a routine task you do regularly

2. Map your current process (step by step)

3. Design your AI-assisted version (where AI helps)

4. Document the workflow (so you can repeat it)

5. Test it (run through the process once)

6. Refine (based on what you learned)

---

Deliverables

Workflow Documentation:

(Use the template from Step 6 above)

Include:

Workflow name and purpose

Time commitment estimate

Risk level

Each step (what you do, what AI does, verification)

Reusable prompts

Verification checklist

---

AI Coach: Workflows Are Where AI Becomes Professional

Here's the gap between "using AI" and "being good at AI":

Using AI: "I asked ChatGPT to draft this email"

Professional AI use: "I run this through my customer response workflow: AI drafts based on template, I verify for accuracy and personalization, then I send"

The difference is systems thinking. You're not using AI ad-hoc. You're building processes that make AI reliable and repeatable.

Key insight: The most valuable thing you can do with AI isn't a single amazing prompt. It's a repeatable workflow that saves time and ensures quality.

---

Reflection: Your Workflow Potential

What routine task could be transformed by AI?

Reflect:

What takes you 15-30 minutes regularly?

What would you automate if you could?

What's the barrier to automation (judgment? verification? change resistance?)?

If you automated this, what would you do with the freed time?

Write 3-4 sentences about your workflow opportunity.

---

Portfolio Check

Portfolio principle: Your portfolio demonstrates professional AI use.

This means:

Outputs are high quality and verified

You can explain how you created them

You used consistent processes

You understand the trade-offs you made

Document your processes. They're part of your professional credibility.

---

Bonus Challenge

Document someone else's workflow: If you work with colleagues using AI, ask them to walk you through their workflow. What do they do well? What could improve? This teaches you to see systems you might miss in your own work.

---

Key Takeaways

Workflows have 5 components: Input, AI Process, Verification, Refinement, Output

Basic workflow: Input → AI → Output

Standard workflow: Input → AI → Verify → Refine → Output

Design by: choosing routine task, mapping current process, identifying AI opportunities, designing AI steps, adding verification, documenting

Reusable elements: template prompts, verification checklists

Common patterns: Summary→Distribute, Template→Personalize, Intake→Categorize, Data→Analysis

Next: The Foundations Layer is complete. You're ready for Lesson 12 and beyond — Departments, Missions, and advanced applications.

---

Congratulations

You've completed the Foundations Layer of The Admin Lab.

You now understand:

✅ What AI is and how it works

✅ Large Language Models specifically

✅ How to prompt effectively

✅ How to verify output

✅ Why AI fails (hallucinations & limitations)

✅ Responsible and ethical AI use

✅ Security and privacy requirements

✅ When NOT to use AI

✅ How to use conversation and iteration

✅ How to structure outputs

✅ How to build repeatable workflows

You're ready for the next phase:

The Foundations Layer taught thinking. The Departments and Missions teach execution.

In the next phase, you'll apply these principles to specific roles (Communication, Research, Operations, Marketing, Data, Executive Support) and complete real missions that build a job-ready portfolio.

You have the foundation. Now it's time to build professional AI skills through practice.

Welcome to Phase 2: Core Training.