Agentic AI Chapter 11 / 16 · Architectures

11 Agent Architectures

Different problems need
different thinking.

How you structure an agent's reasoning is called its architecture — and knowing these patterns is what separates people who dabble with AI from people who ship it. You already met ReAct in Chapter 4. Here are six shapes of thought, and when to reach for each. Click through them.

ReAct Chain-of-Thought Plan & Execute Tree of Thoughts Reflection LATS

The gallery

Six reasoning shapes.

Each pattern trades simplicity, cost, and power differently. Pick one to see its shape and where it shines.

Think Act Observe
step 1 step 2 step 3 answer
Make a plan execute 1 execute 2 execute 3
best path
Attempt Result Reflect ✗ ✓ done
0.4 0.8 0.2 0.6 0.9 highest-value path wins

reason + act

ReAct

Best for
Cost
Same LLM underneath — you're choosing the shape of its reasoning to fit the problem.
What most production systems actually use

ReAct + a dash of Reflection. Simple enough to be reliable, powerful enough to recover from surprises. Reach for Tree of Thoughts or LATS only when a problem genuinely needs deep exploration — they're far more expensive.

Chapter 11 in one breath

Pick the shape of thought to fit the task.

ReAct loops. Chain-of-Thought reasons in a line. Plan & Execute commits up front. Tree of Thoughts and LATS explore many paths. Reflection learns from failure. Start simple — ReAct plus Reflection covers most of it — and escalate only when the problem earns it. Next: what happens when one agent isn't enough.

🖥 Present this chapter

A matching slide deck with speaker notes — press S for notes, F for fullscreen.

Open the slides →