In this episode of The Missing Manual, JCK and Julia Levy go "under the hood" of their own AI experiments to discuss what happens when the tech fails to meet your standards. From Julia’s frustrating experience cloning her voice to JCK’s "agentic design" approach, they break down why AI is a "one-size-fits-one" solution.

They explore the "Step Zero" of process mapping, the power of handoff documents, and why you should treat different LLMs (Claude vs. ChatGPT vs. Gemini) like specialized team members rather than a single source of truth.

THE BREAKDOWN:

[00:00] Intro: The Missing Manual for TA leaders and practitioners.

[01:06] Julia’s Use Case: Building a Talent Intelligence web app and the frustration of "bad" AI voice clones.

[02:15] Defining "Good": Why AI reverse-engineers the wrong results without established guardrails.

[03:15] The Golden Rule: Don’t use AI to automate a broken process (Stop making "better bad processes").

[04:36] Model Switching: When to use Gemini vs. Claude vs. ChatGPT for specific tasks.

[09:31] The "Learning Out Loud" Community: Curating your own AI news pulse.

[11:18] Beyond Business: How Julia used AI to plan a multi-generational family trip to Moab, Utah.

[15:54] The "Challenge Me" Prompt: How to stop your AI from being a "Yes-Man."

[19:39] Final Summary: The critical role of Handoff Documents in grounding your AI.

Tired of your AI losing the plot mid-conversation? Learn why Agentic AI requires Handoff Documents to prevent hallucinations and maximize your token usage in Claude, Gemini, and ChatGPT.

As we move further into the era of Agentic AI, the biggest hurdle isn't the AI's capability, it's its memory. In this episode of The Missing Manual, Julia Levy and JCK (Jenny Cotie Kangas) break down the "handoff" workflow. We explore how AI weights the end of long conversations more heavily than the initial guardrails, often leading to "side quests" and errors.

Whether you are building custom agents for home improvement or managing complex HTML coding projects, a Handoff Document acts as a strategic SOP for AI. It captures model numbers, root causes, and version history, allowing you to "collect the dots" and restart fresh without losing progress.

What you’ll learn in this episode:

• The technical "hockey stick" effect of AI credit usage.
• How to force an AI to generate its own version-controlled handoff.
• The difference between a "Calibration" and a standard audit.
• Real-world demo: Troubleshooting a gas fireplace using a specialized GPT.

00:00 Intro: The Missing Manual Part 3

01:15 What is a Handoff Document in Agentic AI?

02:30 Why Claude and GPT-4o "forget" your instructions

04:00 Julia’s secret to stopping Gemini hallucinations

06:15 LIVE DEMO: Building a Handoff Doc for home repairs

09:45 The "How would you improve this?" power-prompt

12:30 Organizing your AI Handoffs: Folders vs. Trackers

15:00 Why your business needs AI SOPs (Standard Operating Procedures)

20:00 How to share AI context with your team

21:00 Final Tips: When to hit the "restart" button

In Part 2 of our Agentic AI series, we get tactical. Julia and JCK discuss the Brand Voice Artifact—a project designed to ensure AI never misses your tone, style, or specific "guardrails" again.

We explore how Julia used her book, transcripts, and frameworks to create a custom "Gem" in Gemini that acts as her professional voice clone, and how JCK uses "Usage Guardrails" to push back on the AI when it gets too generic.

What you’ll learn today:

• How to collect 'The Dots' (transcripts, writing samples) for your AI.
• Creating a Brand Artifact vs. a simple prompt.
• The 'Nuance' check: Getting AI to understand your specific audience.
• How to handle usage limits and 'maxing out' Claude/GPT.