Project Management

Why PMBOK vocabulary matters for AI programmes

AI engagements often fail for delivery reasons unrelated to model quality:

Failure	How PM vocabulary helps
"Agile" used to mean no scope baseline	Scope knowledge area + change control
Discovery never ends; no gate to build	Process groups: initiating → planning → gate
Data access "coming soon" with no dependency owner	Integration + risk registers
Security review after build	Quality + risk planning upfront
Benefits workshop once; no baseline	Stakeholder + benefits linkage in charter
Vendor POC becomes production by accident	Procurement + scope validation

The PMBOK® Guide Fifth Edition is not an AI methodology. It is the lingua franca many enterprises use for charters, RAID, change control, status reporting and stage gates. When you speak that language fluently, you reduce friction with PMO, finance and SI partners—and you protect clients from unstructured experimentation dressed as "innovation."

AI-specific nuance: Treat model, prompt, eval dataset and retrieval index as deliverables with version control and change impact—analogous to software configuration items—not as informal lab notes.

Why this category exists

Commercial and AI programmes fail less often when scope, schedule, cost, risk, and stakeholders are managed with a shared language. This category is a study and practice map—original summaries and exercises—not a substitute for PMI's copyrighted publication.

Pair it with:

Commercial & delivery
Delivery & programme management
Stakeholder management
Core deliverables — implementation and transition sections

Start here

PMBOK® Guide Fifth Edition — study map — chapters 1–13, process groups, and how the knowledge areas connect.
Pair with Learning Map Stage 5 and Delivery & programme management.
For engagement politics and RACI, use Stakeholder management.
Blog deep-dive: PMBOK® Guide Fifth Edition — practitioner study guide.

Process groups mapped to AI delivery

PMBOK Fifth Edition defines five process groups (not project phases—they repeat). Map them to a typical AI solution lifecycle:

Process group	AI programme meaning	Typical AI artefacts
Initiating	Qualify opportunity; name sponsor; authorise discovery	Opportunity brief, charter, stakeholder register, initial RAID
Planning	Define scope, schedule, cost, quality, risk, comms, procurement	Requirements catalogue, WBS, integrated plan, risk register, eval strategy, architecture baseline
Executing	Build, integrate, train, test, conduct change activities	Sprint delivery, integration builds, training runs, POC demos
Monitoring & controlling	Track variance; control changes; replan	Status reports, burn-up, change log, eval regression gates, FinOps dashboards
Closing	Accept deliverables; KT; benefits handover	Acceptance sign-off, KT plan, benefits tracker, lessons learned

AI gate example (Monitoring & controlling)

Before promoting a model/prompt change to production, run a mini change control:

Impact on eval metrics (link to AI Evaluation)
Security review if tools or data scope changed (Security)
Updated release notes and runbook (Operations)
Approver recorded (RAI / architecture forum)

This is perform Integrated Change Control in PMBOK terms—not optional agility.

Ten knowledge areas mapped to AI Solution Engineering

Knowledge area	AI programme focus	Learning Map deepen
Integration	Charter, integrated plan, change control, lessons learned	25, 8
Scope	Problem boundary, requirements, WBS, acceptance criteria	4, 6, Core deliverables
Time	Phases, dependencies (data, APIs), critical path to go-live	25, 24
Cost	Build + run TCO, tokens, contingency	7, 22
Quality	Eval strategy, NFRs, test strategy—not only "accuracy"	21, 15
Human resources	Team shape, SI mix, KT roles	33, 25
Communications	Steering packs, exec readouts, status rhythm	28, 27
Risk	RAID, threat model linkage, model risk	17, 19, 25
Procurement	Build-buy-partner, SOW, vendor eval	30, 31
Stakeholder	Power/interest, engagement plans	27, 26

You do not need to perform every process personally. You must know which artefact satisfies which process and challenge gaps in client PMO maturity.

What you will practise (artefact table)

Focus	Typical artefacts	AI-specific notes
Integration	Charter, integrated plan, change control log	Charter links value hypothesis to scope; change log includes prompt/model versions
Scope	Requirements, WBS, baseline, validation	WBS includes data pipeline, eval, integration, change workstreams—not only "model"
Time & cost	Schedule model, milestone plan, EVM-style tracking	Include data remediation and security gates on critical path; token run-rate in cost baseline
Quality	Quality plan, eval gates, test strategy	Quality = functional + eval + safety + UX acceptance
People & stakeholders	RACI, communication plan, engagement matrix	Name economic buyer, technical gatekeeper, sceptics
Procurement	Make-or-buy, SOW, contract control	POC exit criteria; IP/data processing terms; model licence

Deepened artefact guidance

Project charter (Initiating) — Should answer: purpose, measurable objectives, high-level requirements, assumptions, constraints, summary budget, named sponsor, PM authority, success criteria. For AI: explicit non-goals ("no production without eval gate X").

Work breakdown structure (Planning) — Decompose into deliverable-oriented work packages: data engineering, retrieval/index, application, integration, security controls, eval harness, change management, hypercare. Avoid WBS that ends at "Fine-tune LLM."

Risk register (Planning → M&C) — Include AI-specific risks: data licence failure, eval drift, prompt injection incident, vendor API deprecation, organisational resistance, unrealistic automation of regulated decisions. Each risk: ID, description, category, probability, impact, response strategy, owner, status.

Stakeholder engagement assessment (Planning) — Current vs desired engagement level per stakeholder; actions to move blockers. Pair with Stakeholder management.

Procurement SOW (Planning → Executing) — Deliverables testable against acceptance criteria; data access responsibilities on client; exit/handback; security and privacy schedules.

How to study with the PMBOK Fifth Edition page

Use PMBOK® Guide Fifth Edition — study map as your spine, not a single reading session.

Recommended 8-week study path (4–6 hours/week)

Week	Study map focus	Practice on live/fictional AI case
1	Ch 1–2: project definition, organisational context	Write one-page charter for your case
2	Ch 3: process groups overview	Map your case to five process groups with dates
3	Integration + Scope knowledge areas	Draft requirements + WBS (2 levels deep)
4	Time + Cost	Milestone schedule + order-of-magnitude cost (build + run)
5	Quality + Risk	Eval gate definition + risk register (10 items)
6	HR + Communications	RACI + comms plan for steering monthly
7	Procurement + Stakeholder	Build-buy-partner table + engagement matrix
8	Closing + integration review	Acceptance criteria + lessons learned template

Study habits that work

One official concept → one client artefact each session—avoid highlight-only reading.
Keep a glossary of PMBOK terms you actually used in status email or RAID this week.
Compare firm templates to PMBOK—name gaps (often change control and benefits).
After each chapter block, answer Competency test Q17–Q18 in writing.
Cross-read blog practitioner study guide for exam-style mnemonics if pursuing certification.

What not to do

Do not treat Agile as absence of scope baseline—integrate user stories into WBS and acceptance criteria.
Do not memorise ITTO tables without case application—certification prep is optional; client credibility is mandatory.
Do not copy PMBOK text into client docs—use PMI-licensed materials for exam prep; use your firm's templates for delivery.

AI programme failure modes (delivery lens)

Failure mode	Early signal	PMBOK-aligned response
Eternal discovery	No charter refresh; no go/no-go	Initiating: reconfirm charter; close discovery phase with acceptance
Scope creep via prompts	"Small tweak" bypasses test	M&C: change control with eval regression
Hidden data dependency	Slips multiply silently	Planning: dependency log with owners; escalate in integrated plan
Security late engagement	CISO first sees prod URL	Planning: quality/risk processes upfront; gate before UAT
Demo = production	No readiness gap analysis	Scope validation + closing criteria differ for POC vs prod
Status theatre	Green RAG, no eval metrics	M&C: honest variance reporting; quality metrics in status
Benefits orphan	No owner post go-live	Stakeholder/Integration: benefits owner in charter; closing handover
Vendor lock-in surprise	No exit in SOW	Procurement: contract change control; alternatives in plan

Hybrid delivery: Agile, waterfall and AI spikes

Most AI programmes are hybrid. PMBOK does not mandate waterfall; it mandates planned, controlled delivery.

Practice	PMBOK alignment	AI note
Two-week sprints	Executing process group	Sprint goals map to WBS work packages; definition of done includes eval regression
Stage gates	M&C between phases	Gates align to Core deliverables—qualification, design, prototype, go-live
Spikes / POCs	Planning risk responses	Time-boxed; success/fail criteria in charter; stop rules documented
Product backlog	Scope baseline + change log	Backlog items trace to requirements IDs; scope creep visible in change register

Anti-pattern: "We're agile so we don't need a risk register." Agile teams still need RAID, dependencies and change impact on eval and security—especially when prompts and retrieval configs change daily.

Status reporting template (AI programme)

Use in steering packs—maps to Communications and M&C:

BLUF: On track / at risk / off track for next milestone (one sentence).
Scope: Completed vs planned deliverables this period; change requests approved/pending.
Schedule: Milestone variance; critical path item (often data or integration).
Cost: Burn vs plan; token/run-rate vs forecast (FinOps).
Quality / eval: Metric trend; regressions blocked releases?
Risk / RAID: Top three with owner and action date.
Decisions needed: Explicit asks (Competency test Q22).

Certification vs practice

Goal	Path
Client credibility this quarter	This overview + PMBOK study map + weekly artefacts on live case
PMP® / CAPM® exam	Official PMI materials + ITTO study; use our map as concept index only
AI Solution Engineer capstone	Delivery + Core deliverables implementation section + Q17–Q20 in Competency test

Exam prep rewards memorisation; role readiness rewards artefacts. Do both only if your timeline allows—never substitute flashcards for a RAID log on a real programme.

Sponsor conversation: When an AI programme lacks a charter, offer to draft one in 90 minutes using the integration knowledge area template above—this single artefact often unlocks PMO support, data access and realistic dates.

When you finish the eight-week path, attach your charter, risk register and status template to an internal pattern library entry (Personal Effectiveness) so the next engagement starts from evidence, not blank slides.

Quick reference: who owns what on AI programmes

Concern	Primary owner on client	ASE role
Charter and scope baseline	Sponsor + PMO	Author or co-author; ensure value hypothesis in charter
Eval quality gates	Product / risk / QA	Define metrics and thresholds (Evaluation)
Security sign-off	CISO / InfoSec	Threat model and control evidence (Security)
Benefits realisation	Business owner	Benefits tracker design (Commercial)
Change adoption	Change lead / BU	Comms and training plan input (Change)

You orchestrate—you do not replace these owners—but you fail the programme if their artefacts are missing from the integrated plan.

Delivery & programme management — AI-specific delivery patterns
Stakeholder management — Power, interest, politics
Business Learning — Stage 1 commercial artefacts
Core deliverables — Implementation and transition sections
Competency test — Q17–Q18, Q20
Blog: PMBOK® Guide Fifth Edition — practitioner study guide
Blog: Communication — how to speak with confidence

Attribution and disclaimer

PMBOK®, PMP®, CAPM®, and related marks are registered trademarks of the Project Management Institute, Inc. Content in this Learning Map category is an independent educational summary for study and workplace practice. It does not reproduce the text of A Guide to the Project Management Body of Knowledge (PMBOK® Guide) – Fifth Edition (© Project Management Institute, Inc.). For authoritative definitions, process details, and exam preparation, obtain the official guide and current PMI materials from PMI.org.