Skip to main content

Tying it all together

📖 Lesson content

What you'll learn

Estimated time: 50 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 workflow augmented with AI
  • Identify which repetitive business tasks are good candidates for AI automation and which should remain human-led
  • Construct clear Product, Process, and Performance Descriptions to guide an AI-powered workflow
  • Plan for Creation, Deployment, and Transparency Diligence when automating business processes

Tying it all together

(10 minutes)

This video brings together all four dimensions of the 4D Framework to build a practical workflow automation for common small business tasks.

Key takeaways

  • Start with Problem Awareness: analyze your actual workload — what are customers asking, and what patterns repeat?
  • Task Delegation means asking "should AI do this?" not just "can it?" — documented FAQs are great candidates; complaints and judgment calls stay human.
  • Build with rich Description: define what the system produces, the step-by-step logic, and the tone, boundaries, and behavior rules.
  • Test iteratively with real examples to practice Discernment — real-world testing reveals gaps your descriptions miss.
  • Practice Diligence: review outputs before they reach customers, be honest about AI's role, and provide a clear path to a human.

Exercise

Exercise 1: Mapping automation opportunities

Identify which repetitive tasks in your business 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 — responding to customer inquiries, updating social media, generating invoices, following up with leads, compiling reports. For each task, note:

  • How often does this task occur? (Daily, weekly, monthly)
  • How long does it take each time?
  • Is the response or process mostly standardized, or does it vary significantly case to case?
  • What's the consequence if this task is done imperfectly?

Part II: Categorize by AI-appropriateness

Sort your tasks into three categories:

  • AI can handle: standardized responses, documented information, clear repeatable processes — e.g., answering FAQs about hours, return policy, or service offerings.
  • AI can assist, human decides: tasks where AI can draft, prepare, or organize, but a human reviews before anything is sent — e.g., drafting personalized follow-up emails or summarizing customer feedback.
  • Human should handle: high-stakes decisions, emotionally sensitive situations, or complex judgment — e.g., resolving a complaint, negotiating a contract, or handling a service failure.

Part III: Prioritize

Choose one task from your "AI can handle" or "AI can assist" categories that would save you the most time if automated. Write a brief sentence about why this task is your top priority — consider time saved, frequency, and how straightforward the automation would be.

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?
  • Did the potential consequences of errors influence your categorization? How?

Exercise 2: Building your automation Description

Describe an automation system using the three types of Description from the 4D Framework.

Part I: Define your Product

For the task you prioritized in Exercise 1, write a clear description in 2–3 sentences:

  • What is the end result you want? (e.g., "A drafted email response to common customer questions about our services.")
  • What inputs will the system receive? (e.g., incoming emails, a customer inquiry form, a list of FAQs)
  • What outputs should it produce? (e.g., a categorized email with a draft response, a flagged item for human review)

Part II: Define your Process

Outline the step-by-step logic — like writing instructions for a very capable but very literal new employee:

  • What should the system do first when it receives an input?
  • What decision points exist? (e.g., "If the question is about pricing, use the attached price list. If it involves a custom request, flag for human review.")
  • When should it escalate to a human?
  • What information or documents does it need access to? (e.g., your FAQ page, price list, cancellation policy)

Part III: Define the Performance

Define how the system should behave:

  • What tone should it use? (friendly and professional, casual and warm, formal and concise)
  • How should it handle uncertainty? (e.g., "If you're not sure about the answer, say so and flag for review — never guess.")
  • What should it never do? (e.g., "Never promise a specific appointment time. Never share another customer's information. Never offer a discount without approval.")
  • How should it acknowledge the person's request before responding?

Part IV: Test with real examples

Share your three descriptions with an AI tool along with 3–5 real examples from your business (past customer emails, inquiries, etc.). Evaluate the outputs:

  • Did it categorize correctly?
  • Are the responses accurate and appropriate in tone?
  • Did it fabricate any information it didn't actually have?
  • What adjustments do your descriptions need based on what you observed?

Iteration is normal and expected — plan to go through at least 2–3 rounds of refinement before the system feels reliable.

Exercise 3: Planning for Diligence

Think through the responsibility and safety aspects of your automation before you deploy it.

Part I: Creation Diligence

Answer honestly about your planned automation:

  • Why is this specific task appropriate for AI to handle? What characteristics make it a good fit?
  • What could go wrong? List at least three potential failure modes (e.g., AI gives incorrect pricing, responds to a complaint with a templated answer, shares information meant for a different customer). For each, how would you catch it? What safeguard can you put in place?
  • What's the real-world impact on your customers or business if AI makes a mistake on this task?

Part II: Deployment Diligence

Plan your review process:

  • Will you review every output before it goes to a customer, or sample periodically? Why?
  • If you're reviewing everything initially, at what point would you feel comfortable reducing oversight? What evidence would you need?
  • How will you monitor for problems over time? (weekly spot-checks, customer feedback tracking, error logs)
  • What specific triggers would cause you to pause or shut down the automation entirely?

Part III: Transparency Diligence

Decide on your transparency approach:

  • Who needs to know AI is involved? Consider customers, employees, and business partners.
  • How will you disclose AI's role? Write the actual language you would use (e.g., a footer on emails, a note on your website, a mention in your terms of service).
  • What follow-up options will you provide if someone wants to speak with a real person?
  • Does your industry or local jurisdiction have any specific requirements around AI disclosure? If you're unsure, this is worth researching before deploying.

Putting it all together: Write a brief "Automation Readiness Statement" — 3–5 sentences summarizing what you're automating, why it's appropriate, what safeguards you have in place, and how you'll be transparent about it. This can serve as a reference document for yourself and anyone else on your team.

Lesson reflection

  • How did using all four dimensions together change your approach compared to earlier lessons?
  • Was describing an automation precisely enough for AI to execute easier or harder than you expected?
  • After the Description and Diligence exercises, is your Exercise 1 task still the right one to automate first?
  • How has your thinking about efficiency versus personal customer relationships evolved?

What's next

In the next lesson, we'll discuss strategies for integrating AI into your business thoughtfully and sustainably — including how to build an AI use policy for your team and how to manage concerns about AI dependency.

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, plus any feedback you may have. Share your feedback here.

🎬 Video transcript

Source video: Fpmd9Z7bpHs

No transcript available — see lesson body for narrative content.

🔁 Related lessons

📚 Source & attribution

Feedback / ReportSpotted an issue or have an improvement idea?