Four roles.
One AI-capable organization.

Everyone in the organization. All roles, all levels. Prerequisite for every other stage.

Make people competent, non-wasteful users of tools they are already using, usually badly. Not experts, not enthusiasts. Competent.

45 min, 60 min, 90 min, half-day, full day. Every format ends with the system prompt workshop.

What AI actually is (the nested hierarchy, hallucinations, knowledge cutoffs, the context window, data hygiene) · basic prompting (zero-shot, few-shot, RAG) · why model output varies · advanced techniques (Tree of Thought, Stepwise Prompting, Metaprompting) · reading and stress-testing output · prompt frameworks (RTF, RODES, Gravity, Onion, and building your own) · iteration as a habit · system prompts.

• A working mental model of what a language model does: enough to predict its failure modes rather than be surprised by them
• The ability to write structured prompts and to know when a framework isn't needed
• Calibrated professional skepticism: four or more tells that an output deserves scrutiny, and three or more ways to check it
• The habit of treating a first output as a draft
• A personal system prompt they built, tested, and intend to keep using

Internal champions: the strategically positioned people who sit close enough to the work to know where the time actually goes. Not IT. Not executives.

AI initiatives fail at the logistics layer, not the technology layer. Ambassadors are trained in logistics: finding the real opportunities, scoring them honestly, building the structures that survive contact with the organization, and representing the organization during the build.

The portfolio is the assessment. Artifacts are checked for named humans rather than abstract roles, feasibility scored against current state, observable escalation triggers, and a KPI baseline captured before deployment.

Why AI projects fail · writing an AI vision statement · stakeholder mapping (Influence-Interest, RACI, resistance) · opportunity identification and the Priority Matrix · governance (data policy, decision rights, human-in-the-loop, escalation) · KPIs and measurement infrastructure · the PACE Plan · overseeing the build.

• An AI vision statement a colleague who wasn't in the room can explain accurately
• A stakeholder map with RACI and a named-blocker communication plan
• A Priority Matrix-scored opportunity list with a defended top three
• A one-page governance document a stakeholder outside the cohort can follow
• One to two KPIs with measurement infrastructure and a threshold that triggers governance review
• A full PACE Plan (Primary, Alternate, Contingency, Emergency) stress-tested with Gravity
• The question discipline and feedback discipline to serve as an informed client during a solution build

Operational staff who will run deployed agentic solutions day to day. Prerequisite: Literacy.

The Coordinator is the semantic layer between machines and the organization. A dashboard shows whether a system is running. The Coordinator determines whether it is behaving as authorized, whether its handoffs are clean, whether what it produces is worth remembering, and when a human needs to take the decision back.

Two days core plus a third day of client-specific tool work, or 1–2 hours weekly over 4–8 weeks. Coordinators train on the actual solutions their organization deployed.

Systems orientation and reading a Solution Passport · intent verification and scope drift · inter-agent coordination and chain tracing · producing institutional memory · escalation and degraded-mode operations · trust calibration.

A graded practical exam. The participant manages a portfolio of interconnected systems through a compressed operating period while the facilitator injects failures without announcing them: scope drift, chain failures originating two links upstream, security events, outages. Over-escalation and under-escalation are both graded down.

• A documented behavioral baseline for every solution they manage
• An intent verification checklist: what scope looks like, what drift looks like, at what cadence
• A chain map for every multi-system workflow, with the highest-risk seams monitored
• Structured observations written to a standard that will still be useful to an Archivist six months later
• Escalation documentation the recipient can act on without a follow-up question
• A Trust Maturity Registry with rationale and review schedule for every system under oversight

Senior operational staff who will own institutional knowledge and supervise Coordinators. Prerequisite: Coordinator, non-negotiable. Supervisors must have done the work.

Most organizations treat AI projects the way they treat any project: the work gets done, the team moves on, the knowledge evaporates, the next initiative starts from scratch. The Archivist breaks that pattern. They are not a knowledge manager, not a data analyst, not an IT administrator. They are the function that converts operational experience into reusable intelligence.

Organized for AI, not just for humans. A knowledge base built for browsing fails at retrieval.

Second brain architecture (taxonomy, tagging schema, inclusion criteria, platform evaluation) · data quality governance and the intake gate · data flow management, input baselines, and data lineage · Coordinator supervision · knowledge compounding strategy.

New solutions start with context instead of from zero. New Coordinators onboard with months of institutional history instead of blind. Ambassadors see patterns across solutions that no single Coordinator could report.

A three-part portfolio review conducted by Upscaile: second brain design validity, quality judgment, and supervisory competence. Scored MET / NOT MET by two independent evaluators. All three must pass.

• A second brain design a second contributor can apply without asking them: taxonomy, tagging schema, inclusion and exclusion criteria
• Quality standards specific enough that two reviewers reach the same verdict
• Data lineage records naming every input's source, owner, last-validated date, and compliance status
• A three-cadence supervision model and feedback that changes behavior
• The three compounding workflows: solution-scoping query support, Coordinator onboarding packages, and the quarterly Knowledge Brief