PSM-AI study plan

Define core AI vocabulary (model, training data, inference, token, embedding, prompt, context window) in your own words without reference materials

Distinguish predictive/discriminative systems from generative systems and give two Scrum-relevant examples of each

Explain in plain language why a large language model can produce fluent but false output

Map five AI capability categories onto the Scrum Master accountabilities of coaching, facilitation, teaching, mentoring and servant leadership

Evaluate three AI-generated outputs for factual accuracy, relevance and bias using a rubric you define

Identify the mechanisms behind at least five failure modes you observe (hallucination, stale knowledge, over-smoothing, prompt sensitivity, false confidence)

Run a controlled prompt iteration experiment on a backlog refinement task, changing one variable per attempt

Write prompts that supply role, context, constraints and output format, and justify each element

Draft AI-assisted designs for each Scrum event and mark where a human decision must remain non-delegable

Use an AI tool to cluster and theme raw retrospective input, then verify the themes against the source data

Rewrite a Definition of Done and a refinement checklist to address quality checks for AI-generated work items

Compare two AI tools on cost, data handling, integration and output reliability for a single Scrum Master use case

Select two Scrum Master measures (for example flow efficiency or event effectiveness) and define how AI assistance would be judged against a baseline

Analyse a simulated impediment with an AI tool, then critique the output against systems-thinking principles

Prepare and rehearse a coaching conversation in which a team over-relies on an AI tool, and record what you would say

Draft the outline of a 60-minute team session on using AI inside the Scrum framework

Facilitate a team session that builds shared vocabulary and surfaces existing AI use in the team

Co-create a team working agreement covering disclosure, verification duties, prohibited data and review of AI-generated artifacts

Coach two team members to write and critique their own prompts for a real team task

Design safeguards that let team members admit AI errors without blame

Map stakeholders affected by an organizational AI adoption and record their concerns and influence

Express an AI adoption effort as an ordered product backlog with measurable outcomes and thin slices

Define at least three empirical measures that would show adoption is or is not delivering value

Identify policy, procurement and compliance gaps that block safe team-level experimentation

Build a risk register covering bias, privacy, intellectual property, security, over-reliance, transparency and environmental cost

Apply a pre-use review checklist to two real AI use cases and record the go, no-go or mitigate decision

Script an escalation path for a suspected data leak or harmful AI output, naming roles and timeboxes

Connect each identified risk to a Scrum Value and state the behaviour that mitigates it

Complete a timed full-length practice set and tag every miss by domain and by cause

Re-teach your three weakest concepts aloud in under five minutes each without notes

Re-run two earlier deliverables and correct the errors your new knowledge reveals

Consolidate a one-page decision aid for when to use, verify or refuse AI assistance in Scrum work