AIPM-2 Exam Syllabus

AIPM-2 Certification Overview

The AIPM-2 Exam Syllabus defines the Version 1.3 assessment standard for AIPM-2: APD Certified Advanced AI Product Manager, the advanced professional practice level in the APD Certified AI Product Manager pathway.

AIPM-2 certifies Advanced Professional Practice. The practitioner independently owns an integrated AI-native product and can lead a major product area or bounded high-complexity product with appropriate specialist and executive support. This level requires advanced judgment across multiple systems, teams, stakeholder groups, risks, and commercial constraints.

AIPM-2 candidates integrate discovery, product strategy, human-AI workflows, requirements, system and data choices, evaluation, governance, resilient operation, commercialization, adoption, and value realization. They create reusable decision mechanisms and influence others while remaining accountable for product outcomes.

AIPM-2 is based on the APD AI-Native Product Manager Competency Model v3.1 and is aligned with the APD Certified AI Product Manager Pathway Overview v1.2.

AIPM-2 Level Boundaries

What Changed in the AIPM-2 Exam Syllabus v1.3

Version 1.3 replaces the separately submitted integrated case and professional discussion with APD-supplied scenarios, artifacts, and linked objective-item sets. It removes the old component weights, practical passing gates, and case resubmission requirements.

Version 1.3 specifically:

• replaces the case-and-discussion route with standardized selected-response tasks assessed against APD-supplied product materials;
• removes component weighting and practical pass gates — no case mark, course grade, or practical resubmission is combined with the objective result;
• retains level names, complexity boundaries, the eight domain identifiers, target weights, and all source learning objectives;
• separates capability-development targets from direct examination observations, and adds an explicit evidence boundary;
• keeps the form at 80 questions in 120 minutes, with 60 single-select and 20 multiple-response items;
• introduces eight APD-supplied linked testlets of four questions each, with testlet-level analysis and controlled initial release monitoring.

AIPM-2 Product Ownership and Scope

AIPM-2 Product Complexity Classes

• C1 — Bounded: a low-complexity product with clear users, tasks, authority, and limited integrations.
• C2 — Integrated: a medium-complexity product with multiple capabilities or integrations inside a defined product boundary.
• C3 — Critical or Multi-System: high complexity arising from consequence, autonomy, system dependencies, scale, ambiguity, or organizational reach.
• C4 — Strategic System: a platform, portfolio, or organization-level product system with systemic effects.

Complexity is assessed across consequence of error; model and agent complexity; authority to act; data sensitivity and rights; integrations; scale and reliability; domain and regulatory constraints; and organizational scope. A small product can be high complexity. Review each dimension and any overriding high-consequence condition; do not classify by user count, budget, team size, or number of agents alone.

What an AIPM-2 Candidate Owns

AIPM-2 independently owns C2 products end to end. It also supports leadership of a major product area or bounded C3 product with specialist and executive support. Candidates must resolve ambiguous trade-offs, manage dependencies and failure propagation, coordinate multiple workstreams, and connect operational performance to adoption and value.

Typical AIPM-2 work includes an enterprise assistant integrated with multiple knowledge and transaction systems, an agentic workflow with controlled external actions, a multi-segment AI application, or a major platform/API product area. Supplied examination materials must provide enough technical and domain information for a product-level decision; specialist knowledge cannot be assumed without being stated.

AIPM-2 Scope Boundaries

Organization-wide portfolio direction, enterprise risk appetite, cross-business platform strategy, and product leadership-system design are principally AIPM-3 responsibilities.

AIPM-2 can own complex product requirements, pricing and packaging, evaluation systems, and product-level threat analysis within its mandate. Model training, infrastructure implementation, penetration testing, and legal opinions remain specialist responsibilities.

AIPM-2 Candidate Profile and Entry

AIPM-1 or equivalent professional capability is strongly recommended, but certification at AIPM-1 is not strictly required. There is no formal eligibility gate. Candidates must already be able to demonstrate AIPM-1 outcomes, including full bounded-product ownership, rather than only terminology knowledge.

AIPM-2 is appropriate for:

• experienced AI product managers;
• technical product managers working on AI systems;
• agent and workflow product managers;
• enterprise and vertical AI product managers;
• platform product-area owners;
• consultants accountable for advanced product decisions.

Candidates should have meaningful product-management or equivalent delivery experience and inspectable evidence of decisions, failures, and outcomes.

Use AI Product Case Practice to establish readiness. A project portfolio can support preparation, but an employment portfolio is not a prerequisite and is not an examination submission. Candidates with limited product exposure should build AIPM-1 capability before attempting this level. Coding and machine learning engineering credentials are not required.

AIPM-2 Assessment Format and Passing Rules

AIPM-2 uses a single assessed component and a single pass rule. Both the objective result and the pass decision are reported separately from any course or preparation activity: no case mark, course grade, component weighting, or practical resubmission is combined with the result.

Marking is rule-based automatic scoring. Each selected-response question is worth one point; a multiple-response item earns its point only when every correct option and no incorrect option is selected. There is no partial credit, no negative item score, and no single-question veto or separate domain minimum. Results are shown at submission, subject to disclosed integrity or administration holds, and a failing score is never rounded up to create a pass.

APD supplies the examination materials: product briefs; interview or baseline extracts; workflow, state, and authority maps; behavior requirements and acceptance criteria; model/provider comparisons; source and memory policies; API and trajectory records; evaluation cases and results; risk/control logs; launch and operations plans; adoption and outcome-economics tables. Setup time is excluded. No production implementation, public deployment, personal case, portfolio, or live defense is required.

AIPM-2 Domain Blueprint

Exam-domain identifiers are retained for traceability. Domain titles and objectives are updated for v3.1 capability. Exam D1 and competency-model D1 are separate namespaces; the mapping section provides the correspondence. Learning objective IDs take the form AIPM2-D1.1 for the first numbered objective in Domain 1.

The following counts define a standard scored form. Target weights express assessment emphasis, not teaching hours; fixed item counts govern a form, and displayed percentages are rounded. Each question has one primary domain.

Forms must use the stated counts. Domain, cognitive level, task family, and response format are independent dimensions on the same questions; their totals must not be added together.

AIPM-2 Cognitive and Question Design

AIPM-2 Cognitive Level Distribution

AIPM-2 Primary Task Family

AIPM-2 Response Format

Response-format allocation is 60 single-select questions (75%) and 20 multiple-response questions (25%). Testlet sets may share a context, but each question must be answerable without correctly answering an earlier question, and later questions must not reveal earlier keys. Domain and task-family assignment follow the primary decision tested; a testlet item is classified by its primary decision even when it interprets an artifact. Difficulty comes from interacting constraints, evidence, and defensible trade-offs, not unexplained jargon or missing information.

AIPM-2 Exam Domains and Learning Objectives

Domain 1: Opportunity Qualification, Product Strategy, and Value Framing

Weight: 12% | Questions: 10

Domain Purpose

Shape product direction under capability uncertainty and competing value constraints.

Learning Objectives

Candidates must be able to:

1. qualify opportunities using customer/workflow evidence, alternatives, capability limitations, strategic fit, and measurable value;
2. define product strategy, target segments, differentiation, behavior assumptions, and a testable adoption/value hypothesis;
3. compare invest, experiment, partner, reuse, defer, and stop options under uncertainty in models, data, integration, and demand;
4. evaluate product-level business cases and unit economics, including human review, failed outcomes, support, and operating constraints;
5. prioritize roadmap options using evidence, dependencies, reversibility, learning value, and risk rather than feature volume;
6. adapt product direction to a capability release, competitor change, cohort result, or cost shock and communicate the decision to senior stakeholders.

Key Topics

• capability foresight;
• product strategy;
• customer segmentation;
• discovery evidence;
• build-buy-partner;
• value hypothesis;
• competitive response;
• real-option reasoning.

How the Revised Examination Obtains Evidence

APD supplies relevant product materials. Candidates select, interpret, diagnose or evaluate: Strategy and opportunity brief; decision/options memo; value model; evidence-based roadmap.
The Blueprint and Evidence Map links each objective to the material, requested response, competent indicator, scoring route and evidence limitation. “Write,” “build,” “coach” and similar source verbs are not silently converted into claims of observed execution.

Scope Boundary

Product strategy and economics are required; enterprise investment allocation and organization-wide portfolio governance are AIPM-3 scope. No independent construction of a complete artifact or live delivery is directly assessed. The exam tests the related product decisions using supplied evidence.

Domain 2: Human-AI and Agentic Workflow Design

Weight: 15% | Questions: 12

Domain Purpose

Design integrated human-AI and bounded multi-agent workflows with controlled authority and failure recovery.

Learning Objectives

Candidates must be able to:

1. model tasks, actors, decisions, state, system dependencies, handoffs, and consequences across an integrated workflow;
2. choose deterministic steps, model-assisted decisions, a single agent, or bounded multi-agent coordination from task and risk evidence;
3. specify identities, delegated authority, least-privilege tools, consent, approval binding, step/spend budgets, and stop conditions;
4. design retries, timeouts, idempotency requirements, compensation, human escalation, interruption, and recovery across partial failures;
5. resolve human workload, autonomy, auditability, speed, and trust trade-offs without assigning unbounded authority;
6. use trajectory and end-to-end evidence to diagnose coordination failures and redesign the workflow.

Key Topics

• stateful workflows;
• bounded multi-agent systems;
• tool contracts;
• authorization;
• replay and duplicate actions;
• compensation;
• trajectory evidence;
• human control.

How the Revised Examination Obtains Evidence

APD supplies relevant product materials. Candidates select, interpret, diagnose or evaluate: Integrated workflow/state map; authority matrix; exception/recovery specification; workflow change decision.
The Blueprint and Evidence Map links each objective to the material, requested response, competent indicator, scoring route and evidence limitation. “Write,” “build,” “coach” and similar source verbs are not silently converted into claims of observed execution.

Scope Boundary

Advanced product-level orchestration decisions are assessed; implementing agent frameworks or setting enterprise-wide authority policy is not. No independent construction of a complete artifact or live delivery is directly assessed. The exam tests the related product decisions using supplied evidence.

Domain 3: AI-Native Requirements and Complex Product Experience

Weight: 14% | Questions: 11

Domain Purpose

Own coherent requirements and trustworthy experience across complex product journeys.

Learning Objectives

Candidates must be able to:

1. translate strategy and workflows into integrated behavior, data, interface, service, and quality requirements;
2. specify acceptance criteria for uncertainty, citations, refusal, human approval, recovery, latency, cost, accessibility, and misuse;
3. design interaction across conversational, multimodal, embedded, or agentic experiences, including asynchronous progress and handoffs;
4. validate calibrated trust and appropriate reliance across user groups, including affected non-users and high-impact decisions;
5. separate durable product constraints from model, prompt, retrieval, memory, and interface hypotheses that need experimentation;
6. reconcile conflicting user, business, engineering, legal, security, and operational needs and maintain traceability through changes.

Key Topics

• AI-native PRD;
• behavior contracts;
• requirements traceability;
• trust;
• consent;
• multimodal UX;
• delegated action;
• accessibility;
• human review workload.

How the Revised Examination Obtains Evidence

APD supplies relevant product materials. Candidates select, interpret, diagnose or evaluate: Coherent requirements package; interaction/control specification; acceptance criteria; evidence-linked change record.
The Blueprint and Evidence Map links each objective to the material, requested response, competent indicator, scoring route and evidence limitation. “Write,” “build,” “coach” and similar source verbs are not silently converted into claims of observed execution.

Scope Boundary

Complete product requirements ownership is in scope; detailed visual design and enterprise design-system governance are not required. No independent construction of a complete artifact or live delivery is directly assessed. The exam tests the related product decisions using supplied evidence.

Domain 4: Solution Architecture, Model Strategy, and Technical Trade-Offs

Weight: 13% | Questions: 10

Domain Purpose

Lead product-level architecture, model sourcing, and change decisions across a compound system.

Learning Objectives

Candidates must be able to:

1. compare prompting, retrieval, fine-tuning options, rules, tool use, and agent architectures against product needs and evidence;
2. evaluate provider and model choices, open-weight and hosted options, build-buy-partner choices, and operational dependencies;
3. specify behavior and service requirements for routing, fallback, caching, interoperability, versioning, and migration;
4. calculate and compare end-to-end quality, latency, reliability, and cost trade-offs, including unsuccessful tasks and human review;
5. use representative evaluation evidence to assess research/model capability claims and product release readiness;
6. diagnose failures from API exchanges, traces, and dependency maps, then justify a product-level architecture or release change.

Key Topics

• compound systems;
• routing;
• model lifecycle;
• capability evaluation;
• API contracts;
• portability;
• vendor concentration;
• cost/latency budgets;
• change impact.

How the Revised Examination Obtains Evidence

APD supplies relevant product materials. Candidates select, interpret, diagnose or evaluate: Architecture options memo; model/provider scorecard; routing/fallback requirements; migration or capability-release decision.
The Blueprint and Evidence Map links each objective to the material, requested response, competent indicator, scoring route and evidence limitation. “Write,” “build,” “coach” and similar source verbs are not silently converted into claims of observed execution.

Scope Boundary

Product architecture decisions are required; training algorithms, infrastructure implementation, and coding proficiency are not. No independent construction of a complete artifact or live delivery is directly assessed. The exam tests the related product decisions using supplied evidence.

Domain 5: Data, Knowledge, Retrieval, Memory, and Context Systems

Weight: 10% | Questions: 8

Domain Purpose

Lead reliable information and memory behavior across sources, users, and lifecycle changes.

Learning Objectives

Candidates must be able to:

1. define source, rights, lineage, freshness, ownership, retention, residency, and permission requirements across systems;
2. compare retrieval/context approaches using supplied relevance, coverage, grounding, latency, and access evidence;
3. specify memory scope, consent, isolation, correction, deletion, and propagation of permission changes;
4. diagnose cross-source conflicts, stale information, retrieval failure, context contamination, and unauthorized data paths;
5. design feedback and evaluation data flows with quality review, privacy controls, provenance, and contamination prevention;
6. define change-impact and regression requirements for data, knowledge, retrieval, context, memory, and access-policy updates.

Key Topics

• cross-source retrieval;
• context quality;
• memory governance;
• tenant isolation;
• provenance;
• rights;
• evaluation-data flywheels;
• feedback selection bias.

How the Revised Examination Obtains Evidence

APD supplies relevant product materials. Candidates select, interpret, diagnose or evaluate: Data/context architecture brief; access/memory policy; source change plan; feedback/evaluation data design.
The Blueprint and Evidence Map links each objective to the material, requested response, competent indicator, scoring route and evidence limitation. “Write,” “build,” “coach” and similar source verbs are not silently converted into claims of observed execution.

Scope Boundary

Requirements, trade-offs, and evidence are assessed; database administration and low-level pipeline engineering are not. No independent construction of a complete artifact or live delivery is directly assessed. The exam tests the related product decisions using supplied evidence.

Domain 6: Evaluation, Experimentation, and Product Intelligence

Weight: 16% | Questions: 13

Domain Purpose

Establish layered evaluation and product intelligence that guide advanced product decisions.

Learning Objectives

Candidates must be able to:

1. define component, retrieval, model, agent-trajectory, workflow, and end-to-end quality dimensions tied to user and business outcomes;
2. construct representative case coverage by cohort, task, consequence, normal behavior, failure, and misuse; protect holdout evidence;
3. specify human rubrics and calibrate automated judges against reference judgments, addressing disagreement and bias;
4. design experiments with a baseline, hypothesis, comparison, success/guardrail metrics, and defensible uncertainty interpretation;
5. set cohort-specific thresholds, regression gates, continuous evaluation, monitoring, drift/change triggers, and escalation decisions;
6. diagnose failure patterns and reconcile contradictory technical, usage, cost, safety, and business signals into a prioritized product decision.

Key Topics

• multi-layer evaluation;
• judge calibration;
• dataset leakage;
• uncertainty;
• experimental comparison;
• cohort analysis;
• trajectory evaluation;
• continuous product intelligence.

How the Revised Examination Obtains Evidence

APD supplies relevant product materials. Candidates select, interpret, diagnose or evaluate: Evaluation architecture and case set; calibration review; experiment design; release decision; product intelligence dashboard specification.
The Blueprint and Evidence Map links each objective to the material, requested response, competent indicator, scoring route and evidence limitation. “Write,” “build,” “coach” and similar source verbs are not silently converted into claims of observed execution.

Scope Boundary

System-level evaluation design is required; research-level benchmarking or advanced statistical derivations are not. No independent construction of a complete artifact or live delivery is directly assessed. The exam tests the related product decisions using supplied evidence.

Domain 7: Responsible AI, Safety, Security, and Product Risk

Weight: 12% | Questions: 10

Domain Purpose

Lead proportionate product risk analysis and responsible decisions across complex dependencies.

Learning Objectives

Candidates must be able to:

1. map harms, threats, affected groups, data exposure, authority paths, and dependency failures across a product system;
2. assess prompt injection, tool misuse, excessive agency, cross-tenant leakage, supply-chain exposure, and insecure output consumption;
3. translate supplied policy/domain obligations into enforceable product controls, evidence, owners, and lifecycle review triggers;
4. specify red-team objectives, misuse scenarios, control validation, and incident exercises with specialists;
5. resolve residual-risk trade-offs within delegated authority and escalate decisions exceeding the product mandate;
6. maintain auditable governance records and adapt controls after incidents, model changes, cohort harms, or workflow expansion.

Key Topics

• product threat modeling;
• AI impact;
• risk/control traceability;
• isolation;
• responsible AI;
• human oversight;
• red-team evidence;
• residual risk;
• incident response.

How the Revised Examination Obtains Evidence

APD supplies relevant product materials. Candidates select, interpret, diagnose or evaluate: System risk/threat model; control and assurance record; incident exercise; risk acceptance or escalation recommendation.
The Blueprint and Evidence Map links each objective to the material, requested response, competent indicator, scoring route and evidence limitation. “Write,” “build,” “coach” and similar source verbs are not silently converted into claims of observed execution.

Scope Boundary

Product-level threat analysis is in scope; security engineering, legal opinions, and enterprise risk-appetite ownership are not. No independent construction of a complete artifact or live delivery is directly assessed. The exam tests the related product decisions using supplied evidence.

Domain 8: Adaptive Delivery, Resilience, Commercialization, and Value Realization

Weight: 8% | Questions: 6

Domain Purpose

Operate, commercialize, and improve integrated products with sustainable performance at scale.

Learning Objectives

Candidates must be able to:

1. plan staged delivery, dependency readiness, launch gates, service objectives, change control, and cross-team operating ownership;
2. specify observability, fallback, rollback, interruption, incident response, provider migration, and retirement requirements;
3. lead pricing, packaging, route-to-market or internal rollout choices using customer value, cost, demand, and risk evidence;
4. plan workflow adoption, customer success, enablement, and measurable value realization across relevant segments;
5. manage scale-related cost, quality, review workload, reliability, and adoption trade-offs through product intelligence;
6. build reusable practices, coach teams, and adjust roadmaps or operating mechanisms after experiments, incidents, and outcome reviews.

Key Topics

• product operations;
• SLOs;
• resilience;
• migration;
• economics;
• pricing;
• packaging;
• adoption;
• retention;
• change management;
• team learning.

How the Revised Examination Obtains Evidence

APD supplies relevant product materials. Candidates select, interpret, diagnose or evaluate: Launch/operations plan; commercialization or internal adoption plan; value scorecard; reusable decision mechanism.
The Blueprint and Evidence Map links each objective to the material, requested response, competent indicator, scoring route and evidence limitation. “Write,” “build,” “coach” and similar source verbs are not silently converted into claims of observed execution.

Scope Boundary

Advanced product commercialization and operating ownership are required; organization-wide transformation and portfolio policy are AIPM-3 scope. No independent construction of a complete artifact or live delivery is directly assessed. The exam tests the related product decisions using supplied evidence.

AIPM-2 Cross-Cutting Capability Threads

The seven v3.1 threads run through domain questions, testlet contexts, and assessment rubrics. They do not add a second set of scored domains.

• Model and Research Productization. Translate model capability and uncertainty into product choices, evidence gates, and change decisions.
• Agent Authority and Runtime Control. Assess identity, delegation, permissions, approval, budgets, stopping, and recovery at the level’s scope.
• Evaluation-Driven Development. Use representative cases, layered evidence, regression, and accountable release decisions throughout the lifecycle.
• AI Product Economics. Use full in-scope costs, successful-outcome denominators, operating capacity, adoption, and value evidence.
• Builder Fluency. Inspect traces, prototypes, and evaluation evidence. Objective assessment does not directly observe building a working product.
• Platform and Ecosystem Fluency. Judge interfaces, dependencies, portability, service commitments, and shared capability at the level’s scope.
• Domain and Regulatory Translation. Translate supplied obligations into controls, evidence, and escalation; unsupplied jurisdiction-specific law is not required.
• AI-Native Product Agility. Use adaptive discovery, staged delivery, continuous evaluation, feedback, and accountable decisions to improve value, quality, safety, cost, and readiness. It is integrated across domains, not an eighth additional thread score. Framework-specific ceremonies, velocity, story points, and branded scaling methods are outside this syllabus.

How the AIPM-2 Exam Syllabus Maps to the Competency Model

The four model clusters and ten competency domains remain the common core. The syllabus groups them into assessable domains at the relevant accountability level. In this table, CM denotes the competency model, and the domain numbers in the right-hand column all refer to that model.

CM-D1 covers strategy and value discovery; D2 commercialization and value realization; D3 human-AI and agentic workflow; D4 model and system decisions; D5 data, knowledge, retrieval, memory, and context; D6 experience, trust, and control; D7 evaluation and intelligence; D8 delivery and lifecycle; D9 responsible AI and governance; and D10 practice, builder fluency, and leadership.

Role-archetype examples may involve applications, agents, models, platforms, enterprise or vertical products, evaluation and governance, growth, or portfolio leadership. Candidates are assessed on the common core at their level; a platform or model example does not require specialist engineering implementation.

APD-Supplied Case Materials and Testlets

The AIPM-2 form contains eight linked testlets of four questions each. A testlet is an APD-supplied context with linked selected-response items: it is examination material supplied by APD, not a case that candidates prepare, upload, or defend. Contexts stay within C2 end-to-end ownership and major product areas or bounded C3 products with specialist and executive support.

Every item must be answerable from the stated materials without getting an earlier answer correct, and later items must not reveal earlier keys. Necessary dates, rights, assumptions, priorities, and required controls are supplied. The candidate may correctly recommend a non-AI approach, a restriction, a pause, or retirement when the evidence warrants it.

Testlet sets are assembled and reviewed as units. A shared context can create statistical dependence even with independent scoring, so the blueprint requires testlet-level analysis and representative contexts rather than one long case dominating the form.

AIPM-2 Passing Candidate Standard

A passing AIPM-2 candidate makes an integrated decision across value, users, architecture, information, evaluation, risk, operation, and commercialization. They diagnose system behavior rather than optimizing an isolated model score, resolve competing constraints, retain accountability through change, and create mechanisms that other teams can use. Their evidence supports the advanced ownership boundary stated in the Product Ownership and Scope section.

AIPM-2 Preparation and Readiness

Candidates preparing for AIPM-2 should focus on:

• reasoning from unfamiliar product evidence rather than memorizing an answer key;
• using AI Product Case Practice with realistic system, business, and governance constraints;
• producing a practice decision pack for your own learning;
• comparing architecture options;
• investigating a failed agent trajectory;
• assessing a judge against human labels;
• reviewing a cohort-specific release gate;
• calculating full task economics;
• rehearsing an incident, migration, or pricing decision.

Experience, simulations, and purposeful artifact creation are useful preparation but do not count toward the examination score. Course attendance or completion does not replace the examination.

Review AIPM-1 outcomes first. Application PMs should deepen system dependencies and operations; technical PMs should deepen adoption and commercial judgment; platform PMs should connect developer experience and service reliability to value; enterprise and vertical PMs should practice domain-rule translation and workflow change. Work with representative evidence and explain trade-offs to both technical and business reviewers.

Representative AIPM-2 Question Patterns and Examples

Examples are illustrative and are not live exam items. Each real question must identify the decision objective, material constraints, and evidence needed for a defensible answer.

Example 1: Mandatory Gates and Full Outcome Economics, Single Select

Four candidate designs meet the overall success threshold but differ in critical-cohort performance, human-review hours, and cost. The stated eligibility rules require both an overall and a critical-cohort success threshold, and the review function has a hard capacity limit. Which decision approach is supported?

A. Choose the design with the highest aggregate success rate, because it is safest overall.

B. Choose the design with the lowest service cost, because review capacity can be expanded later.

C. First exclude designs that fail a mandatory threshold or exceed the review capacity limit, then compare full cost per successful outcome among the remaining designs.

D. Choose the most expensive design, because higher cost implies stronger controls.

Answer: C. Mandatory thresholds and operating capacity are eligibility conditions, not tradeable scores. Only designs that clear every gate are compared on full cost per successful outcome. The most expensive, safest-sounding, or highest aggregate-quality option is not automatically correct.

Example 2: Partial Failure, Single Select

A supplier-order agent times out after sending a purchase request. The external system may have created the order, but the response is missing. What requirement best addresses duplicate-order risk?

A. Retry immediately with a new transaction identifier.

B. Ask the model to infer whether the first request probably succeeded.

C. Increase the retry budget to improve completion.

D. Use a stable idempotency key and reconcile status before any further action; escalate unresolved state.

Answer: D. Completion state must be reconciled under a reliable action contract; a new request or inferred success can create a duplicate or hide a missing order.

Example 3: Release Evidence, Multiple Response

A new model improves average success but weakens a small critical cohort. An automated judge disagrees with domain reviewers on several harmful outputs. Select all justified actions before release.

A. Inspect and calibrate judge disagreements against the defined rubric.

B. Accept the aggregate improvement as sufficient evidence for every cohort.

C. Recheck cohort-specific and critical-risk gates with representative cases.

D. Keep the candidate model restricted or paused if the relevant gates fail.

E. Remove the disputed cases from evaluation to stabilize the score.

Answer: A, C, D. The evidence requires calibration and cohort-aware release judgment. Removing difficult cases or relying on aggregate performance conceals the unresolved problem.

Employer Interpretation of AIPM-2 and Progression

AIPM-2 is a capability signal for advanced individual contributors and product leaders owning integrated or major high-complexity product areas. It is relevant to advanced AI application, agent/workflow, enterprise/vertical, and platform/API PM roles. A company’s “Senior” title may fit this scope; titles and years of service alone do not establish level.

The credential is a standardized signal of assessed product understanding, analysis, and judgment at the retained level. It is not a direct verification that the holder created a full prototype, conducted all required work, influenced actual stakeholders, or independently delivered a successful product.

Employers should inspect independent evidence of work performance, request a failure diagnosis and a defended trade-off, and assess role context, domain experience, and outcome ownership across dependencies.

Progression to AIPM-3

AIPM-3 is appropriate when accountability expands to strategic product direction, shared platforms, portfolios, investment, and organizational capability. AIPM-3 requires AIPM-2 or equivalent evidence and mandatory eligibility review.

AIPM-2 Assessment Integrity, Delivery, and Version Transition

Online Examination

For AIPM-1 and AIPM-2, the assessment is closed reference: no external generative AI, search, notes, or assistance. APD supplies any permitted calculator and necessary extracts. Each selected-response question is worth one point. Multiple-response items say “Select all that apply” without revealing the number of correct options. Credit requires selecting every correct option and no incorrect option; otherwise the item earns zero. There is no negative total score or partial credit.

No case mark, course grade, component weighting, or practical resubmission is combined with the result, and there is no single-question veto or separate domain minimum. A domain breakdown is not a separate certificate, and a failing score is never rounded up to create a pass.

Integrity and Results

Identity and integrity arrangements are disclosed before registration. Results may be held only under published integrity, technical-incident, or review procedures; item-exposure protection does not justify changing a candidate’s scoring rule after the attempt.

Delivery and Fairness

APD must disclose the version, timing, permitted resources, language, scoring rules, identity and integrity arrangements, reasonable accommodations, results target, retake terms, and appeal route before registration. Approved accommodations may change timing or format while preserving the assessed capability. Examination environments must provide equivalent information and access; outages, missing exhibits, and defective scoring are administration issues, not evidence of incompetence.

Retakes and Review

Candidates receive their score and exam result at submission. Reassessment uses equivalent secure tasks, and a defective item is corrected through an auditable correction process. APD publishes applicable fees, scheduling, and appeal deadlines, and publishes any retention period for passed AIPM-3 components separately. No automatic deadline, free retake, or component-retention entitlement is created by this syllabus.

Version Transition

Version 1.3 is effective from 16 September 2026 for candidates booked under this version. Existing certificates retain the rules in force at issuance; the new architecture is not claimed retroactively. The Pathway Overview, model assessment guidance where necessary, item-writing guides, question pools, scoring configuration, candidate communications, and booking pages must be synchronized to v1.3.

Validity

Certification validity is two years. Renewal follows APD’s published policy and may include reassessment, continuing professional development, updated evidence, or an approved bridge assessment.

Summary

The AIPM-2 Exam Syllabus defines the Version 1.3 standard for advanced AI product management practice. AIPM-2 confirms that a candidate can independently own an integrated C2 AI-native product and lead a major product area or bounded C3 product, integrating discovery, product strategy, human-AI and agentic workflow design, AI-native requirements, model and system trade-offs, data and memory governance, layered evaluation, responsible AI risk management, resilient delivery, commercialization, adoption, and value realization.

AIPM-2 is the advanced professional practice level of the APD Certified AI Product Manager pathway and prepares candidates for AIPM-3: Strategic Product Leadership.