Lesson 2 of 6
Mapping the solution space
7 min read
The whole room wants to "add an AI agent." But the problem in front of you is just sorting emails into three buckets. What actually fits?
One problem, many shapes
Before you reach for the biggest model, map the space. Most problems fit one of four shapes: rules (fixed logic), predictive (classify or score), generative (produce text or images), or **agentic** (act in a loop). Each has a different cost, reliability, and failure mode.
The solution shape should match the problem shape. A routing problem is classification; reaching for a generative agent is power the problem doesn't need.
Match the shape to the problem
The fanciest option is rarely the fit. Rules are unbeatable when the logic is clear and stable. A predictive classifier nails routing and scoring. Generative shines when the output is open-ended language or images. Agentic earns its complexity only when the task genuinely needs multi-step action — and pays for it in reliability.
Every step up in power is a step down in predictability. An agent that can do anything can also fail in more ways — reserve it for problems that truly need it.
The shape of it
- —Four shapes: rules, predictive, generative, agentic — rising power, rising cost.
- —Match the solution shape to the problem shape, not to the hype.
- —More power means more ways to fail — use the simplest shape that solves it.
You need to route each incoming email to one of three teams. Which shape fits best?
Continue in the app
Take the whole Discovering & Scoping AI Opportunities course — tracked
Get your personalized path, progress and streaks in the app — this lesson and every next one, in order.