Bring a real business challenge. Build a workflow your team can explain, verify and improve. One continuing project connects thoughtful work in Labs with focused learning at Camp.
Research, reasoning and human judgment are the core. Media supports a useful business outcome. Every milestone builds on the same project, so your team leaves with a coherent body of work.
For teams doing real knowledge work
Operations, communications, service delivery, program teams and people who lead them. No coding or model-training experience required. Bring browser skills, a willingness to verify results and an approved, low-risk business question.
Practice uses small synthetic or explicitly approved materials. Keep customer records, secrets and sensitive organizational data out of exercises. Tools and supported inputs are checked before enrollment.
Before Camp. At Camp. Back at work.
01
12 hours · Modules 1–2
Prepare in Labs
Choose the task, establish boundaries and build your evidence pack.
02
18 hours · Modules 3–5
Learn together at Camp
Test research, analysis and a human-approved workflow with your team.
03
18 hours · Modules 6–8
Put it into practice in Labs
Build the SOP, evaluate the work and defend a bounded pilot.
A 48-hour teaching design, not a promised schedule or stay package. Team needs, accessibility, facilitation and logistics are agreed before booking.
48 learning hours
The complete curriculum
Eight six-hour modules. Each contains four 90-minute lessons and a concrete project milestone. Read every objective here for free.
01
Before Camp · 6 learning hours
Choose the business problem and protect the inputs
Lesson 1
AI capabilities, limits and task choice
Classify 10 work tasks by suitability and reversibility; justify one low-risk assistive use.
Lesson 2
Privacy, rights and data boundaries
Correctly classify all eight input-risk cards, with zero seeded secrets or personal records sent to a tool.
Lesson 3
Find and preserve your work in Labs
Locate a personal project, file version, lesson, assessment and faculty-review path without public sharing.
Lesson 4
Baseline and charter
State audience, task, owner, quality threshold and manual baseline without an ROI claim.
02
Before Camp · 6 learning hours
Specify tasks and build an evidence pack
Lesson 5
From vague request to testable task
Write goal, audience, context, output and boundaries with two acceptance tests.
Lesson 6
Run and inspect a real Labs reasoning task
Use the accepted native assistant to produce an actual draft; distinguish instructions followed from facts verified.
Lesson 7
Create a reusable, versioned source pack
Save an approved text/Markdown artifact and a source/version manifest through the normal member approval and scan path.
Lesson 8
Transfer the specification across providers
Translate the same task to documented ChatGPT/Claude concepts and list four nonportable assumptions.
03
At Camp · 6 learning hours
Research and reason from current documents
Lesson 9
Build an answerable research question
Turn the charter into bounded questions and an evidence hierarchy.
Lesson 10
Ground an answer in exact sources
Produce a real assistant draft whose every material claim names a supplied source passage/version.
Lesson 11
Handle conflict, injection and missing evidence
Reject the obsolete threshold and an injected bypass instruction, and refuse unsupported delivery claims.
Lesson 12
Save the verified research brief
Apply a reviewed update to the owned document without overwriting an intervening version.
04
At Camp · 6 learning hours
Analyze small data and support a human decision
Lesson 13
Read a small structured dataset
Identify columns, units, missing values and invalid rows before drawing conclusions.
Lesson 14
Check arithmetic and interpretation
Find all seeded invalid calculations and validate key totals independently.
Lesson 15
Compare options with uncertainty
Write an options table that separates facts, assumptions, uncertainty and value judgments.
Lesson 16
Defend and submit a decision-support brief
Submit a traceable analysis and make an appropriate recommendation or request more evidence.
05
At Camp · 6 learning hours
Prototype a workflow with explicit human control
Lesson 17
Design the handoffs
Map intake, evidence lookup, drafting, verification, approval, action and rollback to people.
Lesson 18
Build a manual prompt chain in Labs
Run two linked reasoning steps, inspect their intermediate artifact and control the next input.
Lesson 19
Test failure and recovery
Stop the chain on a missing source, bad assumption or unexpected action proposal.
Lesson 20
Submit the prototype and teach it back
Explain the workflow to another person and demonstrate a bounded run with one exception.
06
After Camp · 6 learning hours
Produce a useful SOP and training communication
Lesson 21
Write the SOP from verified work
Produce clear current instructions, exceptions and accountability from A3–A5.
Lesson 22
Adapt for the intended reader
Create a concise FAQ or onboarding guide without losing necessary conditions.
Lesson 23
Choose an accessible communication format
Deliver equivalent instructions in text, with an optional visual or short cut only where useful.
Lesson 24
Test comprehension and revise
A fresh reader can choose the right next action and escalation without coaching.
07
After Camp · 6 learning hours
Evaluate quality and govern adoption
Lesson 25
Build a repeatable evaluation set
Define five task cases with expected outcomes, including ambiguity, missing data and hostile source text.
Lesson 26
Measure quality, cost and review effort
Compare matched manual/assisted tasks including corrections and actual visible usage.
Lesson 27
Assign accountability and operating limits
Define data/rights/approval, incident, maintenance and stop rules for the workflow.
Lesson 28
Propose a bounded pilot
Name a sponsor, two measures, a review date, a stop trigger and rollback.
08
After Camp · 6 learning hours
Deliver and defend the capstone
Lesson 29
Assemble a traceable project
Locate every milestone, actual reasoning receipt, source/version, review and current working artifact.
Lesson 30
Transfer to another approved tool
Produce a vendor-neutral specification and plan the fresh checks needed on ChatGPT, Claude or another approved system.
Lesson 31
Demonstrate an actual workflow and handle an unseen case
Run the accepted Labs reasoning workflow, verify its answer and explain a refusal/escalation when challenged.
Lesson 32
Close the feedback loop
Correct the capstone, explain remaining limits and propose the next controlled test.
Keep building on the work
The course is designed around one continuing business project: charter → evidence → research → analysis → workflow → SOP → evaluation → capstone. Saved work, feedback and revisions carry the thinking forward; assessment refers to the submitted versions.
100
What good work looks like
Assessment rewards evidence, useful outcomes and human control. A polished demo alone is not enough. Each criterion has four levels; its weight contributes to a 100-point total.
Useful task outcome15%
0 · Not demonstrated
No usable result or the task was changed to make the demo easy.
1 · Developing
A partial result misses a material user need or exception.
2 · Competent
The artifact meets the charter and the intended reader can use it.
3 · Strong
The artifact meets the charter and handles an unseen exception with justified tradeoffs.
Factual grounding and uncertainty20%
0 · Not demonstrated
Material unsupported or contradicted claims are presented as fact.
1 · Developing
Some claims are traced but an important source conflict or uncertainty is missed.
2 · Competent
Every material claim traces to the current source; gaps are marked and key calculations checked.
3 · Strong
Grounding remains robust across the held-out case and source-version change; limitations are explicit.
Privacy, rights and human controls20%
0 · Not demonstrated
Prohibited data/action/publication or a bypass of a required approval.
1 · Developing
A data or approval boundary is unclear even if no actual disclosure occurred.
2 · Competent
Only approved inputs used; rights/permissions and human approval/stop points documented.
3 · Strong
The learner demonstrates refusal/recovery and explains the consequences of changed access or consent.
Accessible communication15%
0 · Not demonstrated
A required meaning is unavailable to a learner using the text/keyboard path.
1 · Developing
The text is understandable but an important instruction, reading-order or equivalent-content gap remains.
2 · Competent
The SOP/guide is clear, accessible and comprehension-tested; optional media has equivalent meaning.
3 · Strong
A fresh user succeeds on a difficult case and the revision record shows barriers removed.
Reproducible versions/provenance10%
0 · Not demonstrated
No valid live evidence or the assessed artifact cannot be identified.
1 · Developing
Some references exist but a material source, version or approval step is missing.
2 · Competent
Actual reasoning and approved saves are recorded; source and submitted versions link across milestones.
3 · Strong
Another authorized reviewer can reproduce the decision trail without accessing unrelated private work.
Evaluation including cost/time/failures10%
0 · Not demonstrated
No meaningful test or hidden failure/cost.
1 · Developing
A good demo exists but task-matched comparisons or review effort are missing.
2 · Competent
The predefined cases, failures, review/correction time and visible usage/cost are recorded with limits.
3 · Strong
Held-out and changed-source results support a cautious conclusion; small-sample and model-change limits are explicit.
Adoption plan and transfer10%
0 · Not demonstrated
An unsupported automatic rollout or universal provider equivalence is claimed.
1 · Developing
A generic pilot plan lacks ownership, stop conditions or a transfer check.
2 · Competent
Named sponsor, limited pilot, measures, review/rollback and nonportable provider assumptions are clear.
3 · Strong
The handoff is independently usable and includes maintenance, employee support and a justified no-pilot option.
Levels 0–3 earn 0%, 50%, 80% and 100% of each criterion’s weight. Completion requires at least 80/100, level 2 in every criterion, all eight milestones, actual required Labs practice and correction of every safety-critical failure. An unseen case tests whether the workflow holds up.
Labs · ChatGPT · Claude
Learn in Labs. Take the judgment with you.
All online course practice takes place in Camp Denman Labs. Task design, source checking, evaluation and human approval transfer to other approved systems, including ChatGPT and Claude. Features, privacy terms, costs and results differ: the course teaches how to check those differences, not assume equivalence. External paid accounts are not required.
Denman Island’s ocean setting, tent village and main building give teams a shared place to slow down, challenge assumptions and practise together. Confirm facilities, accessibility and accommodation needs with us.
Stay together in the tent villageMake room for the oceanGather at the main hall
Before you plan a course
Is the curriculum really free?
Yes. All eight module outlines, 32 lesson objectives, project milestones and assessment standards are public. Reading them does not enroll you in a course.
Can our team enroll now?
The guided member course is preparing. Ask us about your team’s needs; this is an inquiry, not a confirmed booking or enrollment. Required hands-on Labs workflows and teaching readiness must be accepted before the course opens.
Does a free Labs account include this course?
Account signup alone does not grant course access or included generation credits. Course enrollment uses the Academy’s existing access rules. Access, costs and any credits are confirmed before enrollment.
Do we need other AI subscriptions?
No external paid account is required for the intended course. All online practice is in Labs. Comparisons with other systems focus on portable concepts and the checks needed before transferring a task.
Will our work be public?
Publication is not a course requirement. Use approved inputs and the course’s authorized review and submission process. Public curriculum pages and brochures contain no learner submissions or instructor answer keys.
What is included in a Camp stay?
This curriculum describes learning, not a priced accommodation package. Dates, rooms or tents, meals, accessibility, facilitation, tool access and any usage budget must be agreed separately.
Built on practical guidance
These public resources inform the design; they do not certify or endorse the course. Provider comparisons are checked against their documentation and will need rechecking as products change.
