Wed Not Wet

Agentic AI Application Storyboard

Stage 1 of 8
Introduction
What is
Agentic AI?

Agentic AI is software that coordinates specialized reasoning, information, and actions to achieve an objective—even when the path changes along the way.

You are about to watch an application set a goal, call specialists, retrieve context, use tools when needed, reject a weak plan, loop, and settle on a better solution.
🎯Objective
🧭Orchestrator
🤝Agents
🔁Loop
A real-world style example
👩🏽‍❤️‍👨🏻 🏛️ 🌩️
A couple plans an outdoor wedding. The forecast turns bad. Can software coordinate weather, venue, budget, food, music, and guest needs to preserve the day?
Introducing
Wed Not Wet
An Agentic AI Application
💍☂️

Wed Not Wet

A wedding-planning application that works to preserve the special day whether it rains or shines.

What You Are About to Experience

Watch the architecture come alive

🎯A shared objective
🧭An orchestrator
🤝Specialized agents
📚Retrievals
🛠️Optional tools
🧠LLM-supported reasoning
🌳Contingent paths
🔁Optimization and looping
Ready? Let’s plan a wedding.
Stage 2

What is agentic AI?

At its core, it is still software. A main application—the orchestrator—coordinates smaller, specialized components called agents.

OrchestratorCoordinates work and checks progress
Agent A

Handles one specialized part of the problem.

Agent B

Handles another specialized part.

Agent C

Returns findings to the orchestrator.

The programming language is not what makes the system agentic. The organization of the work does.
Stage 3

The business problem

The couple wants an outdoor wedding. But rain affects more than location: food, music, photographs, decorations, guests, timing, and cost may all change.

🌦️

Changing condition

The weather forecast changes as the special day approaches.

💒

Desired experience

The ceremony, meal, music, photographs, and celebration should still happen.

🧭

Need for coordination

The application must compare alternatives and recommend the best path.

Stage 4

The optimized state

The wedding takes place somewhere With the planned ceremony, guests, meal, music, photography, décor, and celebration preserved as much as possible.
Wed Not Wet OrchestratorSearches for the best achievable wedding plan
The system is not optimizing for “indoors.” It is optimizing for a successful wedding experience.
Stage 5

The agent team

🌦 Weather Agent

Determines whether the outdoor plan remains viable.

🏛 Venue Agent

Identifies indoor, tented, delayed, or alternate-location options.

🍽 Catering Agent

Preserves meal service and dietary requirements.

🎵 Experience Agent

Protects music, décor, photography, dancing, and guest experience.

💰 Budget Agent

Tests cost, contract, and approval constraints.

📣 Communications Agent

Prepares and sends updates only after a plan is approved.

Each agent contributes specialized reasoning. The orchestrator owns the shared objective.
Stage 6

Retrievals and tools

🧠

Agents

Interpret information, compare options, and return recommendations.

📚

Retrievals

Wedding plan, couple preferences, contracts, guest list, venue rules, and budget records.

🛠️

Tools

Weather service, venue calendar, calculator, email, SMS, maps, and booking systems.

Retrieval provides context. Tools obtain live information or perform actions. Availability does not require use.
Stage 7

Who coordinates whom?

The orchestrator owns the overall objective. It assigns focused work to specialist agents, receives their results, and decides what should happen next.

Wedding Objective Preserve the best possible wedding experience
Orchestrator Keeps the objective, selects agents, evaluates results, and controls the loop
assigns work
receives results
Weather AgentAssesses weather risk
Venue AgentFinds viable locations
Experience AgentTests whether the frills are preserved
The orchestrator coordinates the agents. The agents do not replace the orchestrator or independently redefine the objective.
Stage 8

What does an agent use?

An agent is a specialized application component. To complete its assignment, it may retrieve stored information, call an LLM for reasoning, or use a tool for live data or action.

Retrieval Sources Contracts, preferences, floor plans, guest information, budget records
provides context
🤝

Specialist Agent

Receives a task from the orchestrator, gathers what it needs, performs its focused work, and returns a result.

LLM May help interpret, compare, reason, summarize, or generate language
Tools Weather service, calendar, calculator, email, SMS
Retrieval answers: “What stored information should the agent know?”
The LLM helps answer: “How should this information be interpreted or compared?”
Tools answer: “What live information or action is needed?”
Stage 9

How the relationships form a loop

The orchestrator does not simply call every agent once. It evaluates each returned result and decides whether the current wedding plan is good enough.

1. Orchestrator Assigns a focused task
2. Agent Uses retrieval, an LLM, or tools as needed
3. Result Returns evidence or a candidate plan
The orchestrator evaluates the result.

ACCEPTABLE → continue toward execution
NOT ACCEPTABLE → revise the plan, call another agent, retrieve more information, or loop again
The LLM may help inside the loop, but the application’s orchestrator controls the objective, sequence, and stopping decision.
Case Stage 1

Meet Maya and Jordan

👩🏽‍❤️‍👨🏻

Saturday at Willow Creek Gardens

An outdoor ceremony for 120 guests, followed by dinner, a live band, dancing, photographs, and a sparkler send-off.

Outdoor garden ceremony Live eight-piece band 120 guests $4,000 contingency
Wedding-day forecast
78%
chance of thunderstorms
High risk
2 PMCloudy
4 PMRain
5 PMStorms
7 PMClearing
The ceremony is scheduled for 5 PM—the same time the forecast shows the highest storm risk.
Case Stage 2

The orchestrator gathers evidence

🌦 Weather Agent
🏛 Venue Agent
🎵 Experience Agent
💰 Budget Agent
Live toolWeather service

Thunderstorms are likely from 4:30–6:30 PM, with lightning possible.

RetrievalVenue contract

The indoor ballroom is included, but it must be selected four hours before the ceremony.

RetrievalCouple preferences

The live band and formal first dance are identified as must-have experiences.

RetrievalVenue floor plan

The ballroom fits all guests and dinner tables—but not the eight-piece band with a dance floor.

The obvious answer—move everything indoors—does not fully satisfy the objective.
Case Stage 3

Candidate Plan 1 fails

Rejected by orchestrator

Move the entire wedding into the ballroom

This avoids the storm and preserves the meal, schedule, and guest comfort. However, the band cannot fit while retaining a dance floor.

Safety
Guests
Meal
Live band
Full experience
Objective not achieved.

The orchestrator loops: retrieve more information, call additional agents, and test another arrangement.
Case Stage 4

The system searches for a better path

🏛 Venue Agent
🎵 Experience Agent
🍽 Catering Agent
💰 Budget Agent
RetrievalCovered pavilion plan

A separate covered pavilion can hold the band and dance floor after 7 PM.

Live toolUpdated hourly forecast

Storms should pass by 6:45 PM, with a low rain probability afterward.

RetrievalCatering service plan

Dinner can begin indoors while the pavilion is dried and prepared.

CalculationAdded cost

Extra labor, flooring, and guest guidance cost $2,850—within the contingency budget.

A hybrid plan becomes possible: shelter during the storm, then restore the outdoor-style celebration afterward.
Case Stage 5

Candidate Plan 2 meets the objective

Recommended

Indoor ceremony and dinner; covered-pavilion band and dancing

The ceremony begins on time in the ballroom. Dinner follows indoors. After the storm passes, guests move to the covered pavilion for the live band, first dance, cake, and sparkler send-off.

Safety
Schedule
Live band
Budget
Experience
The best plan is neither simply “outside” nor “inside.” It is the plan that best preserves the desired outcome.
Case Stage 6

The approved recommendation becomes action

Wed Not Wet recommendation

  • Move the 5 PM ceremony and dinner into the ballroom.
  • Prepare the covered pavilion for the band and dancing beginning at 7 PM.
  • Use the $2,850 contingency allocation for extra labor and temporary flooring.
  • Send revised instructions to guests, vendors, the wedding party, and transportation staff.
  • Continue weather monitoring; keep the full indoor plan available as a fallback.
📣 Communications Agent
Email guests
Message vendors
Update itinerary
The system stops optimizing once the plan satisfies the objective and the authorized actions are ready to execute.
What We Learned

Agentic AI is coordinated software working toward an objective

🎯

Objective

The application begins with a desired outcome. In Wed Not Wet, the goal is not simply to move indoors—it is to preserve the wedding experience.

🧭

Orchestration

The orchestrator selects agents, evaluates results, determines what is missing, and decides whether to stop or continue.

🔁

Optimization

The system can loop through multiple candidate plans until it finds one that satisfies the objective within acceptable constraints.

🤝

Specialized Agents

Different agents contribute focused reasoning about weather, venue, catering, experience, budget, and communications.

📚

Retrievals and Tools

Agents retrieve stored context and may use live tools, but tool availability does not mean every tool must be used.

🧠

The LLM

The LLM is not the whole application. It is a reasoning and language capability used by the orchestrator or agents where helpful.

Assumptions & Caveats

What this example intentionally simplifies

🎓

Educational Example

Wed Not Wet is designed to introduce agentic AI concepts. Real-world systems are typically more complex and domain-specific.

⚙️

Not an Implementation

This lesson focuses on the architecture and flow of an agentic application rather than programming languages, frameworks, prompts, or code.

🛡️

Production Systems

Operational systems often include approvals, security controls, audit logging, permissions, monitoring, deterministic validation, and governance requirements beyond this introductory example.

The goal of this lesson is to understand the conceptual composition of an agentic AI application. The technical details of building one come next.
Conclusion

Wed Not Wet

A simple wedding case reveals the core structure of an agentic AI application: a fixed objective, specialized agents, retrieved context, optional tools, LLM-supported reasoning, contingent paths, and a loop that continues until a satisfactory outcome is found.

The objective stays the same.
The path changes as the system learns.
💍☀️
The wedding happened.
The experience was preserved.
Interactive Assessment

Check Your Understanding

Answer all five questions, then select Submit Assessment.

1. What is the primary objective of Wed Not Wet?

2. What is the role of the orchestrator?

3. Which statement best describes retrievals?

4. Why did the first indoor plan fail?

5. What is the role of the LLM in this application?

Live Orchestration

Watch Wed Not Wet work the problem

Wed Not Wet Mission Control

Current optimization run

Live
Current goalPreserve wedding experience
Current candidateNot yet selected
Confidence22%
StatusGathering evidence
1
Weather Agent
Checking storm timing and lightning risk.
2
Venue Agent
Reviewing ballroom and pavilion capacity.
3
Experience Agent
Testing must-have features against the candidate plan.
4
Budget Agent
Checking contingency limits and added costs.
5
Orchestrator
Evaluating whether the objective is satisfied.
What learners should notice

The system does not follow one fixed path

The orchestrator calls different agents as new information appears. It may reject a candidate, loop, retrieve more context, and test a better plan before stopping.

Dynamic sequencing Intermediate evaluation Confidence changes Candidate rejection Loop until acceptable
Next Lesson
How is Agentic AI built?

In the next lesson, we will examine the technical composition of an agentic AI application: orchestrator logic, agent instructions, retrieval pipelines, tools, model calls, memory, state, controls, and stopping conditions.

Up next: Under the Hood of Agentic AI