Training that leaves a working system behind
Every session ends with something built. A keynote that shows the whole ladder in ninety minutes, an eight-week core that climbs it one rung a week, four hands-on labs for the agentic layer, and a library for everything that did not fit. The core is vertical-neutral; a swap layer makes it yours.
- 01 Prompt Brief a model the way you would brief a new analyst: role, context, task, format, what to flag.
- 02 Context Load the workspace first. Instructions, reference files, terminology, a never-say list.
- 03 Automate Pipelines, routing and follow-up that run the same way every time and fail loudly.
- 04 Agent Multi-step work handed to a system with permissions, approval gates and a place output lands.
- 05 Orchestrate Supervisor and worker patterns, human gates, and the order to introduce them.
- 06 Build From a description to a working internal tool, and the discipline to change it without breaking it.
The keynote sells it, the eight weeks build it, the labs go deep, the library holds the rest
| Keynote, 90 minutes | Touches all six rungs with a working demo on each. Sets the vocabulary and shows what the eight weeks build. A 60-minute cut keeps both live builds. |
|---|---|
| Core, eight weeks | Ninety minutes a week, in person or on Zoom. Same shape every week: teach, live demo, breakout, report back, what is new this week, one assignment. |
| Deep-dive labs, 2 to 4 hours | Cowork Lab, Claude Code Lab, Vibe Coding Lab, Frontier Agents Lab. Hands-on, with a stated prerequisite and a stated outcome. Sold on their own or bundled as a track. |
| Online library | The six modules cut from the live schedule, plus setup clinics and a compliance corner per audience, so live weeks never bog down in tech support. |
The eight weeks, one line each
Each week says which rung it is on. That is what keeps a nine-technology scope from feeling like a catalog.
| Week 1 · Prompt | AI fluency and the briefing discipline. What models do and where they fail; what never gets pasted. Leaves with three prompts you will use this week. |
|---|---|
| Week 2 · Context | Context engineering. Why the same prompt behaves differently in a loaded workspace. Leaves with a live project that makes later weeks sound like you. |
| Week 3 · Context | Content and creative at scale. A month of material from one brief, edited down. Leaves with a thirty-day calendar with prompts attached. |
| Week 4 · Automate | The automation layer. Pipelines, contacts, data hygiene, workflows, chat and voice, reactivation and reviews. Leaves with a written automation map of your own funnel. |
| Week 5 · Agent | Your first agent. What an agent is versus a chat; setup, plans, permissions, approval gates. Leaves with one real multi-step task completed by an agent instead of by you. |
| Week 6 · Agent | Extending the agent. Connectors, skills, scheduled work, subagents. Leaves with a scheduled or repeatable job running in your own environment. |
| Week 7 · Build | Vibe coding. From description to working software; iteration without breakage; hosting realities. Leaves with a functioning tool you built, shareable by link. |
| Week 8 · Orchestrate / Build | Claude Code and orchestration. Who on your team should have the engineering rung; supervisor and worker patterns; the audit. Leaves with a prioritized ninety-day roadmap you can actually run. |
Deep-dive labs
The agentic depth the ninety-minute weeks cannot carry. Each states its prerequisite and what you leave with.
Cowork Lab, 3 hours
Install and readiness, folder hygiene and permissions, reading a task plan before approving, connectors, your first skill, scheduled work, subagents, and when Cowork is the wrong tool.
- Prerequisite: paid plan and a laptop
- Leaves with two working agents and one scheduled job
Claude Code Lab, 4 hours
Terminal and IDE setup, working a real repository, custom agents, authoring skills, scaffolding an internal tool, review gates, testing, and what not to let it touch.
- Prerequisite: comfort with a terminal helps, not required
- Leaves with a running project
Vibe Coding Lab, 3 hours
Artifacts for internal tools, app builders end to end, smart forms and sites, iteration without breakage, auth, data and hosting realities, deployment.
- Prerequisite: none
- Leaves with a deployed, shareable app
Frontier Agents Lab, 2 hours
Rotating. Self-hosted agent runtimes, messaging-channel agents, the architecture pattern, the security questions, and honest deployment guidance. Rebuilt every cohort.
- Prerequisite: none
- Leaves understanding the pattern, not with production infrastructure
The online library
Self-serve and evergreen. Where the questions go that would otherwise eat live time.
| Prompt engineering deep dive | The full craft, for people who want it. |
|---|---|
| Image, video and creative production | The actual workflow, with tools and cost bands. |
| Search and paid | SEO, answer-engine visibility, ads, retargeting. Channel playbooks. |
| Data, spreadsheets and analysis | The most requested thing nobody schedules time for. |
| Meeting intelligence and knowledge management | Transcripts, notes and institutional memory. |
| Security, governance and AI policy | Acceptable use, vendor review, what to write down. |
| Setup clinics and the compliance corner | Cowork, Claude Code and CRM setup; the swap-in compliance module per audience. |
The same core, cut for the room
To run the program for one audience, five things change and nothing else: who is in the room, the demo documents, the guardrail, the forbidden claim, and the seasonal peak.
Accountants & Bookkeepers
| Who is in the room | A three-partner firm from January to April. |
|---|---|
| Demo documents | Client organizers, prior-year returns, trial balances, K-1s. |
| The guardrail | No AI-authored tax advice. Preparer sign-off stays human. |
| The forbidden claim | Any assurance of a refund or an outcome. |
| The seasonal peak | Build the January campaign in October. |
CAS & Fractional CFOs
| Who is in the room | A recurring-advisory practice with thirty clients and thirty different close processes. |
|---|---|
| Demo documents | Trial balances, bank feeds, board packs, variance reports. |
| The guardrail | The model never invents a driver. The CFO owns every assumption. |
| The forbidden claim | Any forecast presented as fact. |
| The seasonal peak | Year-end close and budget season, built in the third quarter. |
Commercial Lenders & Brokers
| Who is in the room | A credit team keying the same borrower into three systems. |
|---|---|
| Demo documents | Tax returns, bank statements, rent rolls, credit agreements. |
| The guardrail | No model output becomes an eligibility or credit decision. |
| The forbidden claim | Any statement of approval, rate or terms. |
| The seasonal peak | The fiscal year-end pipeline, built in summer. |
SBA Acquisitions
| Who is in the room | A first-time buyer with a signed letter of intent and a clock. |
|---|---|
| Demo documents | Seller financials, interim statements, the loan package checklist. |
| The guardrail | Readiness preparation only. The independent quality-of-earnings report is the lender's to engage. |
| The forbidden claim | Any assurance that a deal will fund. |
| The seasonal peak | The run-up to a rule change, built before the effective date. |
Private Equity
| Who is in the room | A portfolio-company finance team of three reporting to a sponsor monthly. |
|---|---|
| Demo documents | Management reports, KPI packs, diligence request lists. |
| The guardrail | Investment decisions are never automated. |
| The forbidden claim | Any valuation statement. |
| The seasonal peak | The hundred-day plan, built before close. |
How a session runs
Same shape every week, so attendees know what to expect and the material is built once and delivered twice.
| 15 min · Teach | Slides, standing, laptops closed. The idea and where it fails. |
|---|---|
| 20 min · Live demo | On the main display, reasoning narrated. A recorded backup cued for every demo. |
| 25 min · Breakout | Tables of four to six with a printed prompt card. Every breakout completes in a browser on the free tier; if it needs a paid plan or a connector, it is a demo, not a breakout. |
| 15 min · Report back | Two tables present; the room critiques live. |
| 10 min · What is new | Live only, never in the recording. The reason to attend. |
| 5 min · Assignment | One thing to build before next week. |
No tool is taught as a replacement for professional judgment. Every track names what stays human, and the standing rule is to verify the current state of every tool the week before it is taught, because this part of the stack changes faster than any curriculum.
Start with a training day
Tell us the audience and the documents they touch every day. We will propose the keynote, the core or a lab, and the track that fits.