Episode 29 · Season 1
Build Your First Custom AI Agent: A Step-by-Step Walkthrough
AI conversations can feel like alphabet soup, but building a custom AI agent doesn’t have to. In this episode, we break down custom AI agent architecture end-to-end so you can make the right build decisions with confidence.
What you’ll learn
- The key layers of a custom AI agent: UI/interaction → hosting/orchestration → tools & memory → LLMs → monitoring
- The six core tech-stack decisions most teams must make to ship a custom agent
- A concrete example architecture: an agent that triages Zendesk tickets and takes actions in Salesforce
Who it’s for
Decision makers and technical teams evaluating whether to build vs. buy agentic AI.
Guest + credibility
Tim Harting and Gavin Franklin share a practical, implementation-focused breakdown based on how modern teams assemble real agent stacks in the enterprise.
CTA
Explore the resources and companion artifact in the show notes—and share this episode with a leader or team wrestling with AI implementation. (Follow the show for more Agentic AI architecture deep dives.)
Recommended Links
Gavin Franklin on LinkedIn
Tim Harting on LinkedIn
Episode Artifact
Apple Podcasts
Spotify
Youtube
Greg Banks on LinkedIn
Jason Fowler Music
For questions or feedback, please contact us at CirriusTalk@CirriusSolutions.com
We cover
- Why architecture matters
- Why go custom vs platforms
- Start with the interaction layer (UI / Slack / APIs)
- Hosting the agent logic in cloud (AWS/GCP/Azure)
- Tools + integrations (ETL, endpoints, services)
- LLM choice: self-hosted vs API + multi-LLM routing
Chapters
- 0:00Why architecture matters
- 0:28Why go custom vs platforms
- 1:57Start with the interaction layer (UI / Slack / APIs)
- 4:32Hosting the agent logic in cloud (AWS/GCP/Azure)
- 5:52Tools + integrations (ETL, endpoints, services)
- 7:38LLM choice: self-hosted vs API + multi-LLM routing
- 9:18Monitoring and observability (usage, cost, health)
- 10:21The six core decisions overview
- 11:16Example architecture: Zendesk + Salesforce agent stack
- 15:55Flexibility vs complexity + multiple front ends
Resources and downloads
The artifacts Greg’s team built for this episode, free to download.
Build Your First Custom AI Agent: A Step-by-Step Walkthrough
The episode artifact outlines the technical components and a use case for an AI agent architecture developed by Cirrius Solutions, detailing the various components involved in the AI agent’s operation.
Related episodes
Agentic AITop 3 Agentic AI Frameworks (and How to Choose Without Analysis Paralysis)
Agentic AIPrompt Injection: 4 Practical Defenses for Your AI Agents
Agentic AIMCP Servers Deep Dive: Secure, Reusable Tools for AI Agents
The Military MinuteKate Stoker (Merivis) & the Military-to-Tech Path: Structure, Community, PurposeKate Stoker, Merivis
Salesforce at ScaleSalesforce UI Optimization Faster Users, Cleaner Data, Better AdoptionTalk to us about AI agents
Keep going: See how we build AI agents for real businesses, Talk to us about a custom AI agent, More from the Cirrius blog