Eight hours, three working agents.

Registration, coffee, and welcome

Pick up your badge, grab Carolina coffee, and meet the cohort you'll be building alongside. Bring the laptop you actually use at work, signed in to your Google Workspace or Microsoft 365 account.

Welcome & Introductions

Co-founders Mark Hinkle (The AIE Network) and Todd Lewis (All Things Open) open the day with what every attendee will ship before sundown: three working agents and a method for building more on Monday. Ground rules: build first, debate later.

Crack Open the Agent

By the end of this block you can draw an agent on a napkin and explain what each piece does. We cover the three primitives — model, tools, memory — and the agent loop (plan, act, observe, repeat). Live demo: the same task done badly (no tools) and done well (tools + memory), narrated tool call by tool call.

Example workflow: Watch one task — "summarize my unread email and flag anything from a client" — run twice: once with no tools (a confident guess) and once with tools plus memory (it reads the inbox, writes the summary, and remembers your client list for next time).

You walk out with: A working mental model of how agents think and act

Pick Your First Three Agents (S.M.A.R.T. in action)

Before you build anything, you need to know what to build. We use Mark's S.M.A.R.T. framework (Sort, Match, Automate, Refine, Take Control) as the planning tool — starting with the first two steps: Sort your tasks by AI readiness, then Match each one to the right tool — AI, human, or both (human-in-the-loop). You leave this block with three real tasks from your job ready to become agents — and a clear list of work that stays human.

Example workflow: A marketing manager sorts her week, matches "draft first-pass social posts" to AI, "approve the campaign calendar" to Human, and "read the campaign data" to Both — three agents queued before the coffee break.

You walk out with: Three real tasks earmarked to become agents today

Morning break

Refill, regroup, hit the hallway track.

Write Your First Skill

Skills are how you turn a generic agent into one that does YOUR job. We cover the anatomy of a SKILL.md, the trigger-description problem, and two worked examples: (1) the Anti-Slop Skill — a small file that refuses to let slop ship, with banned words, vague-claim detection, and em-dash policing; (2) the Website Audit Skill — a composed skill that orchestrates Lighthouse, SEO checks, and Anti-Slop into one prioritized fix list. Plus the self-improving pattern: every skill keeps a Lessons Learned log it reads before the next run.

Example workflow: A 20-line brand-voice SKILL.md that loads your style guide, bans your no-go words, and quietly checks every draft the agent writes before it ever reaches you.

You walk out with: A working SKILL.md plus the Anti-Slop starter installed

Ship an Autonomous Research Agent (Manus)

Build Your Morning Brief — an autonomous agent that collates three inputs into one daily readout: email triaged by importance, today's calendar with per-meeting prep notes, and the day's industry news ranked by relevance. Manus does the work; you read the result with coffee at 6 AM. We use a workshop test account for the live build, then cover what's safe to wire to real accounts at home vs. what should wait for the Workspace Agents session.

Example workflow: At 6 AM the agent triages your inbox, lists today's meetings with a one-line prep note on each, and surfaces the three news items that actually matter — all in one readout waiting when you sit down.

You walk out with: An autonomous morning brief running on a 6 AM schedule

Working lunch · Three stacks, three tables

Lunch on us. Three tables, three stacks running side-by-side — Claude Cowork, ChatGPT Agents, and Manus — so you can poke at all three with food in front of you and the people who built the patterns within reach.

Put an Agent to Work on Your Data (ChatGPT Workspace Agents)

OpenAI's agentic mode lives where your data and team already live. We wire up Workspace Agents against a real data source you use — Drive, email, or your CRM — define an agent's job and trigger, set up a human checkpoint before anything goes live, and ship it. Real plays we walk through: customer-service triage, sales intel from CRM, internal helpdesk for HR or IT.

Example workflow: A support-triage agent reads each new ticket in your shared inbox, tags it by urgency and topic, drafts a first reply, and stops for your one-click approval before anything sends.

You walk out with: A Workspace Agent running against your inbox, docs, or drive

Afternoon break

Last refuel before the closing block.

Build the Agent That Knows Your Job (Claude Cowork)

Anthropic's desktop agent built for non-developers — local file access, persistent memory, your skills library. We drop in the SKILL.md you stubbed out in the morning, wire up a connector you actually use (Slack, Notion, Gmail), run an end-to-end workflow that touches your real work, install Anti-Slop alongside it so nothing slop-tier ships, and schedule the whole thing as a recurring agent. You walk out with an agent already part of your workflow, not a demo you'll forget by Monday.

Example workflow: You paste a Slack thread, the agent pulls the matching Notion doc, drafts the weekly status update grounded in your real files, runs Anti-Slop over it, and posts it back — on a standing Friday-afternoon schedule.

You walk out with: A personal, file-grounded agent wired to your tools

Turn Drafts You'd Reject Into Output You'd Ship

AI output quality is directly proportional to input quality. We teach the weekly refinement habit that takes an agent from a 3 to a 9: rate today's outputs as used-as-is, edited, or rewrote; for anything you'd rewrite, diagnose why; update your prompt or SKILL.md and append a lesson learned. Refined prompts and skills beat better models — every time.

Example workflow: Friday review: rate the week's outputs, pull the one you rewrote, trace the miss to a single vague line in the prompt, fix the SKILL.md, and log the lesson so next week opens at a 7.

You walk out with: A weekly refinement habit that compounds week over week

Your Monday Plan + Q&A

Mark's Monday-morning punch list — starter agents ranked by ROI — plus quick coverage of governance and what data should never touch an agent. Then open mic for the questions you saved up while building. What stack should we standardize on? How do we govern this internally? How do we explain it to leadership? Bring it.

Example workflow: You leave with a three-line plan: the first agent you build Monday, the data it's allowed to touch, and the one you'll pitch to leadership by Friday.

You walk out with: A ranked punch list of agents to build first this week