The Idea Sprint
Evaluate ideas quickly, identify AI-solvable problems, and avoid building solutions looking for problems.
The Problem with Most Ideas
Here's a pattern I've seen dozens of times: Someone learns about a new AI capability—maybe GPT-4's reasoning, or RAG systems, or multimodal models—and immediately thinks, "I should build something with this!"
They start with the technology. They end up with a solution looking for a problem.
Starting with "What can AI do?" instead of "What problem am I solving?" is the #1 reason AI POCs fail. The technology should serve the problem, not the other way around.
The 5-Question Framework
Before writing a single line of code, run your idea through these five questions. If you can't answer all of them clearly, your idea needs more refinement.
1. What specific problem am I solving?
Not "improving productivity" but "reducing time to find information from 30 minutes to 30 seconds"
2. Who exactly has this problem?
Be specific. "Knowledge workers" is too broad. "PMs managing 5+ projects" is better.
3. How are they solving it today?
Understanding the current workaround tells you what bar you need to clear.
4. Why would AI be better than traditional solutions?
Sometimes a simple script or database query is all you need.
5. What does success look like in 60 days?
A clear goal prevents endless scope creep.
Should You Even Use AI?
This is the question most people skip. AI is powerful, but it's not always the right tool. AI excels at:
- Unstructured Data: Text, images, documents, conversations
- Pattern Recognition: Classification, clustering, anomalies
- Natural Language: Chatbots, summarization, Q&A
For simple CRUD operations or basic data management? A traditional database is probably better.
Write It Down
Here's an exercise before starting any project. Set a timer for 15 minutes and answer:
- Problem in one sentence. No "and" allowed.
- Target user with a name. "Sarah, a PM at a 50-person startup..."
- Three ways they solve it today.
- Why those solutions fall short.
- Your 60-day success metric. One number.
If you struggle to complete this in 15 minutes, that's a signal. Either the problem isn't clear enough, or you're trying to solve multiple problems. That's okay—it means you need more thinking time, not coding time.
Start with the problem, not the technology. AI should be the answer to a question you've already asked.
Be ruthlessly specific. "Improve productivity" is worthless. "Reduce search time from 30 minutes to 30 seconds" is actionable.
AI isn't always the answer. Sometimes a spreadsheet or script is faster, cheaper, and more reliable.
Define success before you start. A clear 60-day goal keeps you focused.