Skip to main content

Workflow augmentation

📖 Lesson content

What you'll learn

Estimated time: 40 minutes

By the end of this lesson you'll be able to:

  • Apply all four dimensions of the 4D Framework together to build a repeatable procedure with AI

Workflow automation

(6 minutes)

This video brings together all four dimensions of the 4D Framework to build a practical workflow automation. You'll follow Emily, a development coordinator preparing for her organization's annual gala, as she sets up an AI-assisted email response system. The video demonstrates how to categorize tasks by what AI should handle versus what needs human attention, how to describe the system's behavior precisely, and how to maintain Diligence through testing and transparency.

Key takeaways

  • Start with Problem Awareness: Before touching any AI tools, analyze your actual workload. What are people asking? What patterns emerge? Make a specific list before deciding what to automate
  • Task Delegation means asking "should AI do this?" not just "can AI do this?": Some tasks (like answering documented questions) are good candidates for automation. Others (like handling complaints or high-stakes requests) should stay with humans
  • Test iteratively with real examples: Use actual emails you've received to test the system. You'll discover gaps in your descriptions that need refinement—this is normal and necessary
  • Practice all three types of Diligence: Creation Diligence means being intentional about what you automate. Deployment Diligence means reviewing outputs before they go out (especially early on). Transparency Diligence means being honest about AI's role, especially if something goes wrong

Exercise 1: Mapping your automation opportunities

This exercise helps you identify which repetitive tasks in your work are good candidates for AI automation.

Part I: Audit your repetitive tasks

Think about your past week of work. List 5-10 tasks that felt repetitive or time-consuming. For each task, note:

  • How often does this occur? (Daily, weekly, monthly)
  • How long does it take each time?
  • Is the response/process mostly standardized, or does it vary significantly?

Part II: Categorize by AI-appropriateness

Sort your tasks into three categories:

  • AI can handle: Standardized responses, documented information, clear processes
  • AI can assist, human decides: Tasks where AI can draft or prepare, but you review before action
  • Human should handle: High-stakes decisions, emotional situations, complex judgment calls

Part III: Prioritize

Choose one task from your "AI can handle" or "AI can assist" categories that would save you the most time. This will be your automation candidate.

Reflection:

  • What criteria helped you decide which category each task belongs in?
  • Were you surprised by how many (or how few) tasks felt appropriate for automation?

Exercise 2: Building your automation description

This exercise walks you through describing an automation system using the three types of Description.

Part I: Define your product

For the task you identified in Exercise 1, write a clear Product Description:

  • What is the end result you want?
  • What inputs will the system receive?
  • What outputs should it produce?

Part II: Define your process

Write a Process Description that outlines the steps:

  • What should the system do first?
  • What decision points exist?
  • When should it escalate to a human?
  • What information does it need access to?

Part III: Define the performance

Write a Performance Description that defines behavior:

  • What tone should it use?
  • How should it handle uncertainty?
  • What should it never do?
  • How should it acknowledge the person's request?

Part IV: Test with real examples

Share your descriptions with AI along with 3-5 real examples from your work. Evaluate the outputs:

  • Did it categorize correctly?
  • Are the responses accurate and appropriate?
  • What adjustments do your descriptions need?

Exercise 3: Planning for Diligence (stretch goal)

This exercise helps you think through the responsibility aspects of your automation.

Part I: Creation Diligence

Answer these questions about your planned automation:

  • Why is this task appropriate for AI to handle?
  • What could go wrong, and how would you catch it?
  • What's the impact if AI makes a mistake?

Part II: Deployment Diligence

Plan your review process:

  • Will you review every output, or sample periodically?
  • How will you monitor for problems over time?
  • What triggers would cause you to pause the automation?

Part III: Transparency Diligence

Decide on your transparency approach:

  • Who needs to know AI is involved?
  • How will you disclose AI's role?
  • What follow-up options will you provide if someone wants human attention?

Lesson reflection

  • How did using all four dimensions of the 4D Framework together change how you approached this automation task?
  • What surprised you about the process of describing an automation system precisely enough for AI to execute it?

What's next

In the next lesson, we'll discuss strategies for integrating AI into your organization thoughtfully and sustainably—from addressing concerns about AI dependency to building an AI policy that reflects your values.

Feedback

As you progress through the course, we'd love to hear from you about how you are using concepts from the course in your work and any feedback you may have. Share your feedback here.

🎬 Video transcript

Source video: vjXso6GR8Sk

📜 Click to expand transcript (cleaned + AI-translated)

Automating Daily Work with the 4D Framework

Let's talk about something that probably sounds familiar: you're drowning in emails. These aren't complicated emails, just the same questions over and over. "What time does the event start?" "How do I change my ticket information?" "Can you send me my receipt?"

For small nonprofits, especially those running events, this becomes a daily, time-consuming reality. Meanwhile, you've got relationships to build, programs to run, and a mission to advance. In this lesson, we're going to use all four dimensions of the 4D framework together to set up automations that reduce the burden of daily work. This is where AI fluency really comes alive: when Delegation, Description, Discernment, and Diligence work together to produce sustainable efficiency gains for your organization.

Scenario: Managing High-Volume Event Emails

Consider this hypothetical scenario: Emily is the development coordinator for a local branch of a cancer research fundraising group. Her annual gala is three weeks out, and every morning brings 20 to 30 emails asking basically the same five questions: donor reports, event details, table assignments, dress code, and parking information.

Some of these she can answer immediately from documented information. Others need human judgment or access to her donor database. Right now, she's spending two hours every morning just responding to emails. That's 10 hours a week that could go towards literally anything else. AI can change that, but she has to do it thoughtfully—not just automating blindly, but using the 4D framework to build something that actually works and doesn't make gala attendees feel like they're being handed off to a machine.

Dimension 1: Task Delegation and Problem Awareness

Before touching any AI tools, Emily needs problem awareness. She should start by looking at her inbox from the past week to identify what people are asking. Then, she makes a specific list for her gala, breaking it down by what should be delegated:

  • AI Delegation:
    • Receipt requests: These use standard templates and AI can be given database access.
    • Event details: Dress code, parking, and schedule are all well-documented.
    • Process questions: "How do I update my ticket information?" has clear, repeatable steps.
  • Human Delegation:
    • Ticket transfers: There is a high risk if this is done incorrectly.
    • Complaints or concerns: These require extra empathy and context.

This is Task Delegation. You're not just asking "Can AI do this?" but "Should AI do this, and if so, how much?" Emily also considers Platform Awareness. This project requires exposing identifying information and funding data, so it is crucial she uses an AI tool that does not train on user data.

Dimension 2: Description and System Build

Next comes the Description phase. Emily needs to define three things clearly:

  1. Product Description: She wants an email monitoring system that reads messages, categorizes them, and either drafts appropriate responses or flags them for additional review.
  2. Process Description: The system should read each email, identify the category, check if it has the information to respond, and then either draft a reply or create a task for human review.
  3. Performance Description: The tone should be professional but warm. The AI must never guess at information; when uncertain, it must flag the email for human review. It should also always address a person's question directly before providing information.

In practice, Emily might use Claude with a Gmail integration, being very specific about each step and testing the responses with real emails she has received.

Dimension 3: Discernment and Iterative Testing

Product Discernment means checking the outputs. Are the email drafts actually helpful? Do they sound like her? Are they accurate?

During testing, Emily observes:

  • Promotional emails are correctly ignored.
  • Seating assignments are correctly flagged for human intervention.
  • Program details and parking information are accurate.
  • Adjustment needed: A response to a shellfish allergy was accurate but felt "weird" because it asked the user to confirm something they had already stated.

This is an iterative process. Emily goes back and adjusts the Performance Description to be more action-oriented. You refine your descriptions based on what you discover; this is how you build a system that actually works.

Dimension 4: Diligence and Responsibility

The final piece is Responsibility. This isn't just about whether the system works technically, but whether it should be used this way at all. Emily practices three types of diligence:

  • Creation Diligence: She has been intentional about which emails AI handles versus which need human attention, preserving donor relationships where they matter most.
  • Deployment Diligence: Initially, she reviews every response and sends it herself. If she chooses to fully automate later, she can do so with confidence because the system has been thoroughly tested.
  • Transparency Diligence: Emily follows her organization's AI policy regarding transparency. She acknowledges that emails are system-generated to save time and provide quicker responses, and she provides clear follow-up options if additional human assistance is needed. This helps tremendously if something goes wrong.

🔁 Related lessons

📚 Source & attribution

Was this lesson helpful?

Feedback / ReportSpotted an issue or have an improvement idea?