#3 How the Chef Builds Your Tasting Menu

An AI agent is a kitchen. You describe what you want — "a light, seafood-focused tasting menu with wine pairing" — and the agent figures out the rest: which tools to connect, which procedures to follow, what to delegate, how to coordinate the team. You don't specify recipes, temperatures, or who handles which station. You describe the outcome. The kitchen delivers it.

That's not a loose analogy. The four systems that make a professional kitchen work map directly onto the four components of an AI agent's architecture.

What Happens After You Order

You tell the server at a fine-dining restaurant: "Something light today — seafood-focused tasting menu, wine pairing." Thirty minutes later, a flawless seven-course dinner arrives. Here's what happened in the kitchen.

The chef checks what's available. Call the fish market — what came in fresh today? Ask the winery about pairing options. Get seasonal vegetables from the farm. One call, one delivery. This is the supply chain. Without it, the chef goes to the market themselves.

The chef pulls the recipes. How to slice the sea bream carpaccio. What temperature for the butter poach. Which plating style to use. A new hire can reproduce the signature risotto just by following the book. This is the recipe system. Without it, every dish depends on the chef's intuition.

The chef delegates to prep cooks. "Dice fifty onions." "Peel the potatoes." "Trim the herbs." Each prep cook does their assigned task and reports back: "Done." They don't talk to each other — they report to the head chef, who checks the result and decides what's next. These are the prep cooks. Without them, the head chef dices onions while trying to cook.

The team finds its rhythm. A seven-course dinner can't be made alone. The grill chef calls out "Wagyu, three minutes!" The saucier responds "Timing the sauce." The pastry chef starts prepping dessert. They work independently but communicate directly to hit every timing mark. This is the brigade. Without it, the sous chef becomes a bottleneck relaying every message.

The Four Systems

Each piece of the kitchen has an AI counterpart.

AI's evolution tracks the same arc. First it was a cookbook — gives you instructions but doesn't cook. Then a meal kit — ingredients and recipe together, but you still cook. Then a solo chef — actually stands at the stove, but handles everything alone. Now it's heading toward a full brigade kitchen: a team of specialists dividing roles to complete a tasting menu.

Things Always Go Wrong

Real kitchens always have surprises. Good kitchens don't try to eliminate them. They build systems to handle them.

The fish market closes unexpectedly — switch to another supplier. Keep multiple supply routes. In AI terms: when an API goes down, route to another MCP server.

Sea bream was ordered but flounder arrives — the chef decides whether to adjust the menu or send it back. Anything from outside needs verification. In AI terms: validate external data before using it.

A prep cook burns the sauce — the head chef reassigns or handles it directly. Always check results, always have a Plan B. In AI terms: when a subagent errors, the main agent retries or reroutes.

Team coordination can break down too. Two chefs reach for the same sauce pot and the flavor gets thrown off — assign clear ownership. The steak is out but the sauce isn't ready, so the steak goes cold — manage sequencing for dependent tasks. Five chefs for a three-item lunch wastes labor — match team size to complexity. In AI terms: separate file ownership, set checkpoints between dependent steps, and don't throw five agents at a three-step task.

The kitchen's playbook for the unexpected carries over directly. The architecture isn't just how things work when everything goes right. It's how things recover when something doesn't.