// slides-5.jsx — Video slides inserted right after Automation #1. // Wide-format screen recordings of the automations running. // Globals used: C, TYPE_SCALE, SPACING, Slide, Chrome. function VideoSlide({ num, total, src, eyebrow, title, clip, clipsTotal = 3, placeholder, label }) { return ( {placeholder ? (
Video {clip} of {clipsTotal}
Drop the clip here.
[ Placeholder — upload the file and I'll embed it, framed exactly like clip 1. ]
) : ( ); } // Three clips that sit right after Automation #1 (the Labor Report). // The arc: manual process → deterministic script → agentic workflow. function S_Video1({ num, total }) { return ( ); } function S_Video2({ num, total }) { return ( ); } function S_Video3({ num, total }) { return ( ); } Object.assign(window, { VideoSlide, S_Video1, S_Video2, S_Video3 }); // ─── Where the deterministic script breaks ─────────────────────────────── // Sits between the deterministic-script video and the agentic workflow. function S_WhereItBreaks({ num, total }) { const breaks = [ { k: 'The format shifts', s: 'A new column, a renamed field, a re-ordered PDF — and the script reads the wrong cell.' }, { k: 'An edge case appears', s: 'A new cost center, a one-off adjustment, a line it has never seen. No rule covers it.' }, { k: 'Something needs judgment', s: 'A number looks off. Is it a typo or a real spike? A script can\'t ask — it just reports it.' }]; return ( A script only does exactly what you told it.} />
That's its strength — and its ceiling. The moment reality drifts from the rules, it breaks.
{breaks.map((b) =>
{b.k}
{b.s}
)}
→
So a human still has to step in.
); } Object.assign(window, { S_WhereItBreaks }); // ─── Transition: What is an agentic workflow? ──────────────────────────── // Sits right after "Where it breaks" (the deterministic script's ceiling) and // before the agentic workflow video. Defines the concept and answers why it // solves the breaking problem. function S_WhatIsAgentic({ num, total }) { const map = [ { broke: 'The format changed', fix: 'Reads the new layout and adapts the mapping itself.' }, { broke: 'An unseen edge case', fix: 'Reasons it through — and asks before it guesses.' }, { broke: 'It needed judgment', fix: 'Weighs the context and makes the call.' }]; return ( So you wrap it in an agent.} />
A deterministic script runs fixed steps. An agentic workflow hands the model a goal instead — and lets it decide the steps, read its own errors, and use judgment. So when reality drifts from the rules, it adapts instead of breaking.
{map.map((m) =>
{m.broke}
→
{m.fix}
)}
↻
Where the script stops cold, the agent keeps going — a human just reviews.
); } Object.assign(window, { S_WhatIsAgentic }); // ─── How I started: took it to Claude (plain English, no code) ─────────── function S_TookItToClaude({ num, total }) { const Arrow = () => (
→
); const Card = ({ label, children }) => (
{label}
{children}
); return ( To Claude.} />
I didn't write a line of code. I handed it the two files and described my job in plain English.
{['The payroll PDF', 'Production Labor.xlsx'].map((f) =>
{f}
)}
“Here's the PDF, here's the spreadsheet. These numbers go here, those go there. Can you do this for me?”
“We can write a Python script to do exactly that.”
); } // ─── A quick primer: what is Python? ───────────────────────────────────── function S_WhatIsPython({ num, total }) { const traits = ['Beginner-friendly', 'Versatile', 'Massive library', 'Standard for AI']; return (
A programming language — beginner-friendly, versatile, and what seems to be the industry standard for AI. In plain terms, it's how you hand a computer a repetitive job — open this file, pull these numbers, write them there, send that email — and have it done perfectly, every time, without you. You don't need to write it yourself; you just need to know it exists, and that Claude can write it for you.
You don't need
To know how to write it.
You do need
To know what it is — and what it can do.
); } Object.assign(window, { S_TookItToClaude, S_WhatIsPython });