Master AI at your own pace

Build a practical map of AI and its neighboring concepts. Learn to tell fixed-rule automation and data analytics from machine learning and generative AI, then describe the model, input, inference, output, and human responsibility inside a real business workflow.

Start with a user or business problem, not a tool. Link one clear task to a result people can see. Record how the work performs now. Then build an Opportunity Card for a small, safe test. Do not promise a return before there is evidence.

Learn how to choose a tool for one clear task. Compare rules, search, prediction, content creation, planning tools, mixed systems, and human work. Then build a Capability Choice Card with the evidence, owner, review step, and smallest safe test.

Learn the main words vendors use when they talk about artificial intelligence (AI). Follow one service assistant from request to answer. Learn what each term means, what can go wrong, and what to ask. Then build a Vocabulary Brief.

Learn seven ways to check a workplace artificial intelligence (AI) tool. Ask if it protects data, works well, treats people fairly, explains its role, and has a clear owner. Then build a Pause Card to help your team continue, fix a problem, or stop.

Choose one sales or marketing task. Link it to a clear moment in the customer's journey. Decide what proof will show value for the customer and the business. You will build a Customer Journey Test Card with a need, task, output, action, result, safety check, and fair comparison.

Design a support path that uses artificial intelligence (AI) without trapping customers in a bot loop. Decide when self-service can help. Set clear routes to a worker. Choose measures for speed, a correct result, access, and trust. Then build a Support Path and Handoff Map.

Map one work process before you choose where artificial intelligence (AI) may help. Decide which tasks need automation, decision support, or a person. Then build a Workflow Control Map. It will show the current result, who handles unusual cases, how people can override the tool, how the team will watch it, and when work must stop.

Learn where artificial intelligence (AI) may assist and where it must not control a serious decision. First, look at help with finance work. Then look at choices about jobs and credit. You will build a small pilot plan with approved data, separate checks, clear reasons, human authority, a way to fix errors, and stop rules. This course is general education, not legal advice.

Learn how to move an artificial intelligence (AI) idea to a new setting without copying the old plan. Compare the goal, people, data, work steps, risks, rules, size, and cost of failure. Then adapt, test, or reject the idea. Your final work is a Use-Case Transfer Test.

Learn how to name the data that an artificial intelligence (AI) use needs. Find where the data comes from, what form it takes, how it moves, who it describes, and what is still unknown. Your final work is a Data Source and Journey Map.

Learn how to test whether data is good enough for one stated artificial intelligence (AI) task. Set a baseline. Write clear pass rules. Check completeness, correctness, consistency, freshness, relevance, coverage, labels, and lineage. Find gaps across groups and situations. Then choose to repair, limit, collect again, or stop. Your final work is a Task Data Quality Test Sheet.

Learn who makes choices about data. See who keeps it useful, runs its tools, checks risk, and may use it. Then map how access is asked for, approved, given, checked, changed, and removed. Your final work is a Data Responsibility and Access Map for one system use.

Build a simple plan to protect personal and business data used by artificial intelligence (AI). Learn how privacy differs from security. Check the purpose, data amount, access, vendor reuse, storage time, deletion, and incident steps. Your final work is a Data Protection Check. This course is for learning. It is not legal advice.

Learn to decide if data is ready for one bounded artificial intelligence (AI) use. Map the data needed to build, test, run, watch, fix, and retire the use. Make a Data Readiness Decision Brief. Choose ready, ready with conditions, discover or repair first, or stop.

Learn to write one clear request for an artificial intelligence (AI) helper. State the work goal, source, reader, output, limits, checks, and review owner. Make a Prompt Brief that helps a person check the result before use.

Learn how to choose one request or split a task into stages for an artificial intelligence (AI) assistant. Set checks before the run. Test common, hard, and failed cases. Compare the result with today's process. Change one key item at a time. Know when to stop or make the task smaller. Build a Task Run and Iteration Sheet.

Learn to use an artificial intelligence (AI) helper for research and writing. Keep control of sources, claims, voice, and final checks. Build an Evidence-to-Draft Pack with a source ledger, claim map, draft brief, and fact check.

Learn when to record, make a transcript, or take short notes with an artificial intelligence (AI) tool. Set data rules. Check the record. Turn each real choice into a clear action.

A tested artificial intelligence (AI) task can become a team method. Learn how to add an owner. Add tests, versions, feedback, rollback, review dates, and clear end rules.

Learn how one artificial intelligence (AI) idea can support a real business goal. Map the change, pick useful measures, test the claim, and name facts that would make the team stop or change its plan.

Learn to compare artificial intelligence (AI) ideas in a fair way. Use clear gates, facts, set scales, unknowns, and trade-offs. Know when to test, score again, pause, or stop. Do not trust one total score.

Learn to shape a small artificial intelligence (AI) portfolio. Connect useful results, shared foundations, risk work, and tests. Match the set to team limits. Make clear stop choices.

Choose how to source one approved artificial intelligence (AI) service. Compare buy, build, partner, open-source, and mixed paths. Weigh fit, control, skill, full cost, proof, dependence, and a safe way out.

Use four earlier work items to make an artificial intelligence (AI) roadmap. Put work in a clear order. Show outcomes, proof, risks, links, and team limits. Add owners, decision dates, date ranges, change rules, and stop choices.

Observe how work happens now. Map the service outcome, start and end, people, steps, choices, handoffs, systems, data, waits, rework, exceptions, controls, and measures. Then choose one small change that is safe to test.

Turn one approved work boundary into a safe no-code flow. Map the trigger, checks, rules, and artificial intelligence (AI) task. Add human review, action, proof, and a record. Set access and test limits. Name owners, stop rules, fallback, and rollback before a live step.

Design human review, approval, override, and appeal for an artificial intelligence (AI) workflow. Put each control where it can change an outcome. Base the choice on harm, doubt, how hard a step is to undo, who has power, and how much work reviewers can handle.

Make a safe error plan for one work flow. Learn how to spot trouble, block risky acts, stop repeats, send work to people, save proof, fix work that stops halfway, and choose when to try again, pause, undo, or stop.

Build a plan to measure one workflow after artificial intelligence (AI) automation begins. Compare a fair starting point with real results. Track service, quality, risk, use, staff and user effects, full costs, and where saved time goes. Then choose to improve, hold, expand, narrow, or stop.

Plan one clear and honest chat for a customer. Say that the helper is automated. Set its limits. Ask only for facts you need. Check key choices before action. Make errors, exit, and human help easy to handle.

Get trusted content ready before an artificial intelligence (AI) helper answers customers. Name the right source and its owner. Set who may use it, where it applies, and when it needs review. Set answer limits and a safe no-answer step. Test that each answer has support. Do not let the helper guess.

Plan one useful and limited form of personalization. Start with a clear customer benefit. Use the least data needed. Prefer a strong default, broad segment, or stated choice before a hidden personal guess. Give people notice and control. Test a small change against a non-personalized path before you expand.

Build a plan for warm handoffs and service repair in an artificial intelligence (AI)-supported journey. Make it easy to ask for a person. Set clear rules for when the tool must stop. Pass only useful, approved facts. Show who owns the case. Help staff act. Restore the customer’s path. Learn from repeat faults.

Build a scorecard. Use it to review one artificial intelligence (AI)-supported journey. Start with the customer result and service promise. Choose a few fair measures. Check reach, task completion, resolution, quality, effort, trust, safety, access, and recovery. Find gaps and work moved to other people. Then choose to improve, hold, expand, narrow, or stop.

Plan one safe link for artificial intelligence (AI). Choose how two systems will connect. Name the source of truth. List the data, access, limits, errors, tests, owners, and backup path. Keep reads apart from acts that change records. Use fake examples. Never enter a key, token, password, private record, or live system fact.

Map one service that uses artificial intelligence (AI). Start with the user need and support route. Show the system edge, parts, data, actions, trust changes, state, time, owners, service signals, weak points, fallback, and recovery. Keep the human service in view.

Build a Pipeline Data Contract, Flow Map, and Failure Runbook for one artificial intelligence (AI) workflow. Move only needed fields from source to sink. Choose batch, event, or stream delivery. Set rules for field names, units, record keys, time, quality, access, and data life. Plan for bad, copied, late, mixed-up, and blocked records. Add safe replay, a hold area, checks, repair, and deletion steps.

Build a security table and stop plan for one artificial intelligence (AI) link. Keep identity checks apart from permission checks. Limit access, secrets, network paths, data, and actions. Check all untrusted input and output. Plan safe logs and alerts. Review vendors and changes. Add a kill switch, response steps, and a safe return to work.

Build an Integrated Pilot Charter, Launch Gate, and Run Plan for one small artificial intelligence (AI) journey. Join earlier contracts and controls. Set clear measures, test the full path, stage access, name owners, and prepare support, rollback, cleanup, and stop rules. Use evidence to expand, improve, narrow, pause, or stop.

Design a clear interaction for generative artificial intelligence. Tell people what the tool can do, what it cannot know, and when a person must decide. Give users choices, previews, and a safe way to correct or stop work.

Turn a generative artificial intelligence task into visible steps. Decide which step may draft, look up, or ask a tool to act. Keep tools narrow. Add checks, approval points, and a safe stop path.

Choose what information a generative artificial intelligence task may use. Write clear instructions. Retrieve only relevant approved material. Set memory limits, source labels, and a safe answer when the context is weak.

Test a generative artificial intelligence application with realistic work cases. Use a clear rubric. Record evidence. Set release gates that say when to improve, limit, pause, or release a small change.

Prepare for a generative artificial intelligence tool to be wrong, unavailable, unsafe, or unclear. Spot failure signs. Protect people from a bad action. Escalate to the right owner. Recover through a safe normal route and learn from the event.

Learn to prepare useful documents for a knowledge assistant. You will name an owner, check source quality, and record changes before information is searched.

Choose meaning-aware retrieval when wording differs but the idea is the same, and show evidence for that choice.

Break information into useful units, keep metadata and permissions with each unit, and update the index when a source changes.

Use representative test cases to measure whether retrieval finds the right evidence and whether answers stay grounded in that evidence.

Design a knowledge assistant that respects permission rules, gathers helpful feedback, and hands people to a human when needed.

Learn to choose an assistant, guided workflow, artificial intelligence agent, or orchestration based on value, harm, and the need for human control.

Define small, permissioned tool actions with validation and clear confirmation before an artificial intelligence agent acts.

Separate task state, approved memory, evidence, and restart boundaries so artificial intelligence agents can recover safely.

Set approval levels, autonomy limits, budgets, stop conditions, and escalation paths for bounded artificial intelligence agents.

Monitor outcomes, trace failures, improve bounded artificial intelligence agents safely, and keep normal workflows available.

Learn to frame supervised machine learning, a method that learns from labelled examples, as a specific prediction task with evidence and limits.

Use exploratory data analysis, a careful first look at information, to form hypotheses about patterns and segments without treating them as facts or decisions.

Design features, input details used for prediction, and labels, known past answers, with representative coverage, provenance, and fairness checks.

Compare predictive models using held-out data, examples saved for testing, suitable metrics, and decision thresholds that reflect real consequences.

Choose predictive machine learning, rules, search, or generative artificial intelligence based on the real job, then define human review that keeps service accountable.

Build proof for one artificial intelligence (AI) release. Record fixed artifact identifiers (IDs) and checksums. Capture the tools, inputs, test results, and sources of random change. Rebuild the release, check rollback fit, and write a report for the next architecture review.

Build an architecture decision record. Use it to show why an artificial intelligence service should use batch, real-time, edge, or hybrid delivery. Use facts about speed, work rate, safe load, where data stays, harm from failure, backup paths, and safe return to service.

Learn how to watch a public service that uses artificial intelligence. Set a starting point, goals, warning limits, and safety floors. Check each group. Study an alert before a control changes, then test if the action helped.

Practice a response to an artificial intelligence incident. The work stays offline. Build a tabletop record. Set severity rules and on-call roles. Record command handoffs and safe containment. Protect the evidence. Plan clear updates and rollback choices. Check the recovery. Own the follow-up work. Do not touch a live system.

Run a small test that cuts cost or saves time without harm to quality or safety. Compare useful results, speed, load, and spend. Set firm stop and rollback rules.

Learn how to turn promises about artificial intelligence (AI) into clear controls. Build a policy control matrix. Name its purpose, owner, proof, approval level, enforcement point, challenge route, end date, and review test. Use the prior measured-constraint pattern only as a template. Create and check a new Atlas Route record for Northstar Public Services.

Learn to judge how much harm an artificial intelligence (AI) service may cause. Map who may be hurt and how. Check facts, gaps, controls, and risk left. Then name who may accept that risk. Keep data, organizational choices, and legal advice apart.

Plan a safe red-team review for an artificial intelligence service. Set clear rules, use test data, log proof, rank findings, fix weak controls, and test them again. Keep all work in a sealed test space. Stop if people or live systems may face harm.

Learn to map data and its uses in an artificial intelligence (AI) service. Build a register for privacy, data rights, contracts, copyright, records, place, sector, and decisions. This is not legal advice. Removing names can lower risk, but people may still be re-identified. Rules depend on jurisdiction, sector, contract, data, and decision type.

Build a reliable evidence pack for a set of different artificial intelligence services. Link each rule to clear proof. Test well-defined groups of records. Answer an outside review. The index helps the review, but it cannot prove that records are true, approved, or complete.

Learn how to test if a clear user problem is worth more work. Build a short evidence brief, check other causes, compare simple options with artificial intelligence, and set a firm gate for the next step.

Turn an approved user problem into clear product rules. Link each rule to a need, owner, test, safe failure path, and proof. Set firm bounds for data, access, risk, and change before any release.

Build an Outcome Evidence Scorecard for a clear product claim. Set a baseline and track outcomes, quality, use, and service effects. Check user groups, guardrails, unknowns, and action limits. Then hand leaders evidence they can test and use.

Run a fair hearing for a hard artificial intelligence choice. Compare real options, evidence, unknowns, limits, and effects on people. Record who may decide, object, pause, or raise a concern. Add an end date so the choice must face review.

Build a small pilot that your team can watch, support, pause, and undo. Set firm limits for each stage. Test the fallback before use starts. Then use fresh evidence to expand, improve, narrow, pause, or stop.

Choose where key work should live from real service facts, not from fashion. Compare speed, shared risk, local needs, rare skills, team capacity, and help routes. Then record clear service lines and review rules for the next team.

This course helps you turn an approved work model into a clear map. You will name who acts, who owns each result, what each role may do, and when staff must seek help.

This course helps you turn fixed role goals into safe work practice. You will build real-life drills, timely help, follow-up tasks, and fair checks that show if staff can act well under stress.

Build a Change Claim & Response Ledger. Use it to match each message to facts, mark doubt, name a human owner, and route hard questions to the right expert.

Build a community service blueprint from the operating model selected in Course 76. Keep central, distributed, or hybrid boundaries unless evidence and an approved review justify a change. Define accountable services, fair service goals, asset rules, exceptions, capacity, and a learning loop.

Learn to map an artificial intelligence service in clear layers. Show interfaces, evidence, failure paths, limits, and named owners. Build a plain system map that helps reviewers find gaps and decide if a design is ready for a closer platform review.

Learn how to choose an artificial intelligence platform with care. Weigh needs, full costs, support, and proof that key work can move. Build a Platform Comparison and Exit Plan with open questions, named owners, clear limits, and a tested path out before the team commits.

Build a clear register for an artificial intelligence service. Map who or what crosses each trust boundary. Set narrow access rules, tests, evidence, and owners. Decide when weak controls must block access or stop a launch.

Build a clear set of limits for a reliable artificial intelligence service. Measure demand, capacity, speed, errors, and recovery. Plan safe backup steps when key parts fail. Give named owners the power to slow, pause, or restore the service.

Create a set of architecture decision records that lets leaders inspect key choices. Compare sound options, link each claim to evidence, record trade-offs and dissent, and show effects on other choices. Add review dates, clear triggers, owners, and reversal costs so old choices do not stay in force without a fresh check.

Turn an approved architecture choice into a business case that leaders can inspect. Measure the current state, count full costs, set benefit ranges, and compare sound options. Test what happens when key assumptions shift. Use value at risk and staged funding gates to limit loss before leaders commit more money.

Turn business claims into clear tests. Define each measure, challenge assumed links, protect groups with harm-based floors, and write decision rules before results arrive in practice.

Use a Portfolio Evidence Board to compare projects and show weak proof. Apply firm gates. Check staff time and links between projects. Use staged funding, learning value, and clear rules to advance, pause, stop, or retire work.

Turn supplier promises into clear duties, proof, remedies, and a tested exit plan. Leaders learn what must be shown before signing, who may accept the risks that remain, and how service, data, and daily work can survive supplier failure or change.

Build a brief that asks for one clear decision. Show the quality of the proof, value ranges, key assumptions, choices, risks, and authority. Turn approval conditions into testable needs for the decision-support system in Course 91 while keeping uncertainty easy to see.

Map how evidence becomes a recommendation and then a human decision. Build an accountable operating picture that shows provenance, assumptions, authority, challenge routes, feedback, fallbacks, and failure points.

Stress-test a decision before rollout. Build a scenario matrix that compares a baseline with controlled changes, interacting assumptions, uncertainty ranges, and credible failure cases without pretending to predict the future.

Compare feasible choices against a clear objective, hard constraints, fairness checks, uncertainty, and competing trade-offs. Build a constrained decision table and a short memo that shows why a responsible choice may not have the highest headline score.

Design a human-oversight operating plan with risk tiers, review thresholds, realistic capacity, escalation, override authority, appeal, and restoration criteria. Make review strong enough to change an outcome instead of becoming a rubber stamp.

Run a post-deployment learning system that links early signals and final outcomes to segmented evidence, drift reviews, careful cause claims, and accountable corrective action. Build a Live Outcome Learning and Drift Register with owners, challenge routes, fallback steps, and stop conditions.

Build a task-specific capability evidence dossier tied to an exact model and version. Use representative tests, a fair baseline, uncertainty, failure slices, total cost, and change control to set limited use boundaries and hand reproducible evidence to later decision stages.

Assure an accessible museum guide by tracing complete user journeys, testing consent and alternatives, examining failure modes, checking assistive technology, and defining accountable human fallback.

Learn to produce a clear ethical decision memorandum for a hard workplace choice. You will compare affected groups, rights, benefits, harms, and practical options; test the quality of the evidence; preserve dissent; name residual risk; and set review triggers that keep leaders accountable after the decision.

Build a strategy for several plausible artificial intelligence futures. Use critical uncertainties, evidence signals, no-regret moves, reversible options, and clear triggers. The course follows North Valley Delivery as it creates a scenario portfolio and trigger map without treating any forecast as certain.

Receive a reversible option from Course 99 and test it as a controlled workplace innovation experiment. Use a falsifiable hypothesis, matched comparisons, safeguards, an evidence ledger, and explicit causal limits to choose a stop, improve, or scale gate.