Systems Thinking and AI Engineering

Track Promise

See how people, information, tools, and decisions work together, and use that understanding to improve AI-supported systems while keeping responsibility, evidence, and consequences visible.

The center of this Track is understanding what changes around AI—not just how the AI works. A faster answer can change who checks information, whose work becomes harder, what gets remembered, and who can act. Start with that surrounding system before choosing tools or adding automation.

This is not only for software engineers. Teachers, creators, community learners, managers, builders, and future consultants can begin with a familiar communication situation. Technical architecture is optional depth, not an entrance requirement.

Start Here

Begin with How Technology Changes the Ecology of Communication — Sketch / Planned, owner Unassigned. It is this Track’s signature module: notice how a new tool changes people’s channels, habits, trust, memory, and decisions.

Its embedded Communication Ecology Map activity introduces the systems view. Choose a low-risk public or fictional situation, such as a community announcement. Map who communicates, through which channels, where trust comes from, and what action follows. Then ask what changes when generative AI joins that situation—and what should remain human-led.

The intended take-home artifact is one page showing the current channels, friction, trust signals, possible AI changes, and human decisions. This is a discussion-and-mapping entry, not a requirement to build or deploy a system.

The sketch is open for feedback. It needs learning-design review and those materials before it can be offered as a complete, self-contained workshop package.

The Systems Loop

Map → Trace → Set boundaries → Test → Observe → Revise

• Map: name the purpose, people, channels, tools, and decision. Include the work that already functions well.
• Trace: follow information from its source through conversations, context, memory, and handoffs. Notice what changes or disappears.
• Set boundaries: name the responsible human, the AI’s limited job, permitted information and actions, and when to stop or ask for help.
• Test: try one low-risk change or compare supplied examples. State what evidence would support or challenge the proposed improvement.
• Observe: look beyond a polished output. Who checks it, who bears the extra work, and what unintended effects appear?
• Revise: keep, change, reduce, or decline the AI contribution. Record the reason, what remains uncertain, and how the system can recover.

A map shows your current understanding, not proof that a system is safe. Keep observations separate from assumptions, AI suggestions, and questions that need outside evidence. More automation is not the assumed destination.

Follow the System

These three routes extend the signature’s map. They are developing directions, not three finished courses or a required sequence. Existing planning sources stay linked for feedback; source-free ideas are labeled. Choose the route that helps explain a question in your map.

Developing route

Follow the information

Follow a source through a conversation, a context packet, saved material, and a handoff. Ask what the next person or tool can see, what it cannot see, and what it may use.

Memory and Retrieval Proposed · no module source

Lab idea: Decompose an AI assistant into its system parts — proposed, no Lab source. Begin with a supplied diagram or paper map, not an installation.

Missing: a reviewed information-flow example, source and handoff labels, a guided Activity, and a standalone Notebook. The shared conversation and context Labs below practice parts of this route; they do not yet teach the whole system.

Facilitator route demonstration: Coming Soon

Developing route

Understand action and authority

Distinguish an AI suggestion from permission to act. Name the decision owner, allowed and prohibited actions, meaningful review, and stopping conditions.

Decision Ownership Before Delegation Seed / Planned · candidate from AI Readiness

Lab idea: Review a system for permissions, memory, observability, and human oversight — proposed, no Lab source.

Missing: a small, reviewed authority scenario and a learner-ready control record. The module’s Decision Ownership Map and Delegation Ladder are candidate artifacts, not supplied worksheets. Review needs enough evidence, time, and authority to say no; a human approval button alone is not a safeguard.

Facilitator route demonstration: Coming Soon

Developing route

Evaluate and improve

Look for feedback, unintended consequences, failure, correction, and recovery. Ask whose experience is absent from a success claim and what would cause you to revise the design.

Designing Systems That Can Admit They Are Wrong Seed / Deferred · candidate from AI Readiness

Observability Proposed · no module source — making the system’s behavior visible enough to inspect and learn from.

Missing: a beginner-friendly failure-and-recovery example, evidence checks, and a guided comparison of the intended result with what happened. The deferred Seed is an ideation placeholder needing research, governance, and learning-design review—not a ready-to-teach design framework.

Facilitator route demonstration: Coming Soon

Strengthen Your Practice

Use these existing module-owned Labs when you want concrete practice with one part of the system. They are Shared Support, not a second AI Literacy core or substitutes for the missing signature package.

• Conversation Engineering — Sketch / Planned. Its Stay, Steer, or Start Over, Beta, practices finding a mismatch, correcting it, challenging the result, and deciding the next move. Open the Lab Notebook.
• Context Engineering — Sketch / Planned. Its Build the Brief Before the Work, Beta, practices selecting context, changing one condition, comparing outputs, and recording a maintenance decision. Open the Lab Notebook.

Both Labs have learner and facilitator materials and standalone Notebooks; learning-design, accessibility, and pilot review remain open. The modules retain their preparation and dependency guidance.

Use the Labs’ decision and comparison records as evidence about a bounded exercise. A chatbot’s explanation does not prove it used every source, understood the goal, or will behave the same way next time. The output-comparison steps need a live operator or an appropriate facilitator-prepared comparison set; sorting a brief alone is not the whole comparison.

Optional: Explore Technical Depth

Follow this branch only when a systems question calls for more technical detail. Coding, runtime setup, tool integration, and deployment are not required for the signature entry.

For later evaluation depth, When the Average Lies: Simpson’s Paradox in AI Evaluation remains Sketch / Deferred. Its imported packet contains synthetic data and separate learner and instructor materials. It is not validated curriculum or a runnable Track sequence; statistical-methods, accessibility, learning-design, and pilot review are still required. Keep instructor answers separate from learner materials.

For adjacent work, visit Building Knowledge Bases / Second Brains, AI Readiness and Risk Assessment Support, or Agentic Coding. These are connections, not prerequisites to mapping a communication situation.

Participation and Access

Paper mapping, discussion, supplied examples, and observing a facilitator are valid routes. No coding, deployment, paid subscription, or learner account is required. You may write, speak, direct a partner, or compare maps without presenting publicly.

A live generative-AI encounter can help test one bounded question, directly or through a partner or facilitator. It is optional here. Record whether your evidence came from a live run, an observed demonstration, or supplied material; do not describe a hypothetical change as a tested result.

The signature’s example packet still needs to be built. For current shared Labs, facilitators must prepare the access or comparison route before delivery; the module’s supplied-example option does not mean every Lab includes a complete no-tool output set.

Use public or fictional examples. Exclude private disputes, customer or student records, credentials, internal incidents, and other people’s sensitive information. Permission to discuss a situation does not imply permission to upload its records. Do not connect live systems or grant execution authority as part of this Track’s beginner practice.

For Facilitators

Begin with the signature sketch and one familiar, low-risk communication situation. Keep the question concrete: what changes for people when AI joins this situation? Use the module’s historical milestones as brief orientation, not a full technology-history lecture.

First build priority: the Communication Ecology Map’s standalone Activity, Notebook, reviewed example packet, and facilitator guide. A short demonstration remains Coming Soon. Do not advertise a complete signature workshop before those gaps are addressed.

While the signature package is developing, the shared Labs offer bounded practice. Use the Stay, Steer, or Start Over facilitator guide or Build the Brief Before the Work facilitator guide, with appropriate preparation and access. A pilot of either Lab tests that practice, not the readiness of the whole Systems Track.

Gather actual timing, unfamiliar language, accessibility needs, handoff confusion, and whether participants can name a responsible person and explain one decision. Keep anonymized observations, not private records or identifiable chat histories.

After the signature materials receive review, pilot one short map: current situation, proposed AI contribution, a boundary, and what still needs evidence. Let that reveal which developing route is most useful next. No new technical course, deployment exercise, or governance framework is included in this pass.

Additional Reading

No additional reading is listed yet.

Status and Feedback

This Track remains Draft / Drafting, owner Unassigned. The signature and shared modules remain Sketch / Planned; the two shared Labs remain Beta. Decision Ownership Before Delegation remains Seed / Planned; Designing Systems That Can Admit They Are Wrong remains Seed / Deferred; When the Average Lies remains Sketch / Deferred.

Source links make drafts available for feedback; they do not establish learner readiness. The signature now names Systems Thinking and AI Engineering as a related Track while retaining AI Literacy as primary and AI Readiness as related. Shared-support placements remain candidates, and the two governance candidates retain AI Readiness metadata. The shared Track Map records the organization and historical audit.

Use this page’s Track feedback and review links for the promise, starting point, route clarity, statuses, access, and composition. Does the communication ecology entry make the systems view approachable? Are the shared Labs useful without becoming the center? What module, Lab, or route is missing, misplaced, or too technical? Teaching details belong in the separate Module and Lab reviews.

This pass creates no module, Lab, Notebook, example packet, or facilitator demonstration.

Team review information. You are viewing the team review site. The contributor review tools below are separate from the learner activity and learner feedback.

Review status

• Package lifecycle: Draft
• Learner availability: Drafting
• Review needs: Needs track-level composition, nontechnical entry, authority boundaries, accessibility, and pilot review.

Review this TrackShare Track feedbackIndependent final QC