AI Immersion · Workshop
Build & Interact
w/ Agents
Please enter a valid key to continue.
AI Immersion · Workshop
Please enter a valid key to continue.
Build & Interact with Agents Workshop · Participant Guide
Today you will move from insight to working Agentic solution. The session is structured as a full design-and-build sprint where you will frame the opportunity, design, specify, and build an agentic solution. Each step builds directly on the last.
Use AI to rapidly frame the problem, architect the solution, and generate buildable specs. Explore the Challenge Card from multiple perspectives, converge on an agentic architecture, and produce precise specifications that become your build inputs.
Opportunity canvas — explore the challenge from multiple angles and generate a structured map of actors, pain points, workarounds, and value at stake.
Solution design & agent decomposition — converge on an agentic architecture: name each agent, define its role, inputs, outputs, orchestration pattern, and human-in-the-loop gates.
Generate & refine specs — produce a precise, buildable specification for each agent (purpose, tools, constraints, success criteria) and review until unambiguous.
Download & save — export all finalised specs to your project folder. These files are the direct input to the build phase — do not skip this step.
Upload your specs directly into Claude Code and generate the working agentic solution. This is where design becomes reality. Once built, run it — test multiple scenarios and observe how your agents behave under different conditions.
Prototype generation — Load your saved spec files into Claude Code and read and run the initiation_prompt present in the project folder.
Solution generation — copy the supporting prompt from handbook and run it in Claude Code.
Run & test — use the files present in the Test Data given to run different scenarios. Observe outputs, edge cases, and handoff behaviour between agents.
Human-in-the-loop check — verify your review gate is functioning as designed. Does it fire at the right moment?
If time allows, explore how your agentic solution would operate at enterprise scale. Use n8n to examine the foundational capabilities that separate a prototype from a production deployment.
Orchestration patterns — how would your agents be coordinated, monitored, and retried in an enterprise workflow engine?
Context & guardrails — where does enterprise context get injected? Where do policy guardrails apply and how are they enforced?
Business evaluation & integrations — what does an evaluation framework look like? How do agents connect to enterprise data sources and systems of record?
Tool Stack