Solving the Right Problems

The biggest reason AI projects fail isn't "Bad Code"; it's "Bad Ideas." Because AI is currently in a hyper-hype cycle, executives often want to "sprinkle some AI" on everything.

As an AWS Certified AI Practitioner, your job is to guide the business toward the High-Impact, High-Feasibility projects. In this lesson, we learn the criteria for a "Good AI Opportunity."

---

1. The Three Filters of a Good AI Project

To identify a high-impact opportunity, a project must pass through three filters:

Filter 1: The "High Friction" Filter

Does the task currently involve a lot of human "Wait time" or "Repetitive labor"?

• Excellent: Reading 5,000 invoices a day and typing them into Excel.
• Bad: Deciding on a company's new logo (This is high-creativity and low-repetition).

Filter 2: The "Data Availability" Filter

Do we actually have the data to solve this?

• AI is not a magician. If you want to predict "Truck maintenance," but you haven't been recording when the trucks break down for the last 5 years, you cannot do AI yet.

Filter 3: The "Acceptable Error" Filter

What happens if the AI gets it wrong?

• Low stakes: A movie recommendation is bad. (High impact, low risk).
• High stakes: An AI wrongly identifies a safe bridge as "Collapsing." (Medium impact, extremely high risk).

---

2. The Feasibility/Impact Matrix

When brainstorming AI ideas, place them on this 4-quadrant map:

quadrantChart
    title AI Opportunity Matrix
    x-axis Low Feasibility --> High Feasibility
    y-axis Low Impact --> High Impact
    "Moonshots": [0.2, 0.8]
    "Quick Wins": [0.8, 0.8]
    "Distractions": [0.8, 0.2]
    "Sinkholes": [0.2, 0.2]

• Quick Wins (Target These First!): Tasks that are easy to do (have data, use managed services) and provide high value (save time, increase revenue). Example: Summarizing customer emails.
• Moonshots: High value, but very hard to do. Example: A cure for cancer using AI.
• Distractions: Easy to do, but no one cares if you do it. Example: Using AI to write the lunch menu.
• Sinkholes: Hard to do and low value. Avoid at all costs.

---

3. High-Impact Patterns by Industry

Industry	High-Impact Opportunity	Correct AWS Service
Retail	Product Recommendations	Amazon Personalize
Finance	Account Opening Fraud	Fraud Detector
Legal	Contract Summarization	Amazon Bedrock (Claude)
Supply Chain	Demand Forecasting	Amazon Forecast
Media	Auto-tagging Videos	Amazon Rekognition

---

4. Summary: The 2026 Strategy

In 2026, the best AI opportunities are those that Augment humans rather than Replace them.

• Instead of "A bot that does legal work," build "A bot that helps a lawyer find information faster."
• This reduces the risk of errors and increases the "Buy-in" from your staff.

---

Exercise: Evaluate the Project

A large hospital wants to use AI to:

1. Automatically transcribe doctor-patient conversations.
2. Flag any mention of "High Blood Pressure" for a follow-up list.
3. Does the hospital have doctors? Yes. Do they have recording equipment? Yes. Is there a managed service for this? Yes (Transcribe Medical).

Where does this sit on the matrix?

• A. Sinkhole.
• B. Distraction.
• C. Moonshot.
• D. Quick Win.

The Answer is D! All the ingredients (Data, Talent, Tools) exist, and the impact (saving doctors hours of typing) is massive.

---

Knowledge Check

What's Next?

Opportunities are great, but how do we prove they were worth the money? In the next lesson, we see how to talk to CFOs about "The Bottom Line." Find out in Lesson 3: Calculating AI ROI and Value.