Knowledge Track
Project Manager
Core frameworks, methodologies, planning tools and governance concepts for enterprise project delivery. Use this as a quick reference before interviews or when refreshing key concepts.
1 Foundations
What a project is, its lifecycle and key constraints.
Project Lifecycle
🚀
InitiationDefine purpose, stakeholders, feasibility
📐
PlanningScope, schedule, budget, risk, resources
⚙️
ExecutionDeliver work according to the plan
📊
MonitoringTrack, review and regulate progress
✅
ClosureFormalise completion and lessons learned
Triple Constraint
Scope
What work is included and excluded. Changes must go through formal change control.
Time
Deadlines, milestones and schedule. Delay on critical path = project delay.
Cost
Budget, resources and financial limits. Tracked via EVM (see section 3).
Project vs Programme vs Portfolio
| Level | Definition |
|---|
| Project | A single temporary endeavour delivering a defined output |
| Programme | A group of related projects managed together to obtain benefits not available from individual management |
| Portfolio | A collection of projects, programmes and operations managed to achieve strategic objectives |
2 Methodologies
Waterfall, Agile, Scrum, SAP Activate, PRINCE2, ITIL.
Waterfall vs Agile
| Dimension | Waterfall | Agile |
|---|
| Approach | Sequential, linear phases | Iterative, incremental sprints |
| Requirements | Fixed upfront | Evolve throughout |
| Change | Expensive, formal | Expected, embraced |
| Delivery | Single release at end | Frequent releases each sprint |
| Best for | Fixed scope, regulatory projects | Complex, evolving products |
Scrum Key Concepts
| Concept | Description |
|---|
| Sprint | Time-boxed iteration (1–4 weeks) delivering a potentially shippable increment |
| Product Backlog | Prioritised list of all work to be done |
| Sprint Backlog | Subset of backlog selected for the current sprint |
| Daily Standup | 15-min sync: what did I do, what will I do, blockers |
| Sprint Review | Demo to stakeholders at end of sprint |
| Retrospective | Team reflection on process improvements |
| Product Owner | Owns backlog and priorities. Represents stakeholders |
| Scrum Master | Facilitates the process and removes impediments |
SAP Activate
🔍
DiscoverValidate scope and business case
🛠
PrepareSet up project and environment
🗺
ExploreFit-to-standard workshops
⚙️
RealiseConfigure and build
🚀
DeployCutover and go-live
🔄
RunOperations and hypercare
3 Earned Value Management (EVM)
The standard method to measure project performance objectively.
Remember: CPI and SPI above 1.0 = good. Below 1.0 = problem. EVM integrates scope, schedule and cost into a single performance picture.
| Term | Meaning |
|---|
| PV (Planned Value) | Budgeted cost of work scheduled at this point in time |
| EV (Earned Value) | Budgeted cost of work actually performed |
| AC (Actual Cost) | Actual cost of work performed to date |
| BAC | Budget at Completion — the total approved budget |
4 Planning Essentials
Scope, schedule, requirements, change control.
Scope Management
- WBS (Work Breakdown Structure): Hierarchical decomposition of all project work into manageable components
- Scope Baseline: Approved scope statement + WBS + WBS dictionary
- Scope Creep: Uncontrolled expansion of scope. Managed through formal change control
- Gold Plating: Adding unrequested features. Equally dangerous — avoid it
- MoSCoW: Must have / Should have / Could have / Won't have — prioritisation method
Schedule Management
- Critical Path Method (CPM): Longest sequence of dependent tasks. Delay here = project delay
- Float / Slack: Time a task can delay without affecting the end date
- Fast Tracking: Tasks in parallel to compress schedule. Increases risk
- Crashing: Adding resources to compress schedule. Increases cost
Change Control
- Change Request (CR): Formal document describing the proposed change
- CCB (Change Control Board): Group responsible for reviewing and approving changes
- Impact Analysis: Always assess effect on scope, schedule, cost and risk before approving
- Golden rule: No change is free. Every change has cost, schedule or risk implications
5 Risk Management
Identify, analyse, respond and monitor project risks.
| Strategy | Description |
|---|
| Avoid | Eliminate the threat by changing the plan |
| Mitigate | Reduce probability or impact to acceptable levels |
| Transfer | Shift risk to a third party (insurance, contract) |
| Accept | Acknowledge the risk and prepare a contingency plan |
| Escalate | Escalate to programme/sponsor if outside PM authority |
Risk vs Issue: A risk is a potential future event. An issue is a risk that has already materialised. Both require an owner and a log entry.
6 Governance & Tools
Roles, meetings, metrics and project management tools.
Governance Roles
| Role | Responsibility |
|---|
| Project Sponsor | Business owner. Approves budget, resolves escalations, champions the project |
| Steering Committee | Senior stakeholders. Strategic decisions, approves major changes |
| Project Manager | Day-to-day management. Owns delivery, risk and stakeholder communication |
| Workstream Lead | Leads a specific track of work |
| PMO | Standards, governance, reporting, tools across the programme |
Key Metrics
SPI / CPI
Schedule and cost performance. Target >1.0
Milestone Adherence
% of milestones delivered on original committed date
Scope Change Rate
Number/value of approved CRs vs original scope
CSAT / NPS
Customer satisfaction and net promoter score
Key Terms
| Term | Definition |
|---|
| RAID | Risks, Assumptions, Issues, Dependencies — review in every status meeting |
| RACI | Responsible, Accountable, Consulted, Informed — defines decision-making roles |
| WBS | Work Breakdown Structure — hierarchical decomposition of all project work |
| UAT | User Acceptance Testing — client testing to validate deliverables |
| Baseline | Approved plan (scope, schedule, cost) against which progress is measured |
| Hypercare | Elevated support period immediately after go-live |
7 Certifications
The main credentials for Project Manager roles.
PMP
PMI. Most globally recognised. 180 questions. Requires 36+ months PM experience. ~$555
PRINCE2
UK-based, widely used in Europe. Foundation + Practitioner levels. Strong in public sector
PMI-ACP
PMI's Agile credential. Covers Scrum, Kanban, Lean, XP
ITIL 4
IT Service Management. Highly relevant for SaaS and technology operations
SAFe
Scaled Agile Framework. Relevant for large enterprise Agile programmes
CSM
Certified Scrum Master — Scrum Alliance. Entry-level Scrum facilitation
Knowledge Track
Engagement Manager
Strategic client leadership, commercial awareness, SaaS delivery and the competencies that distinguish an EM from a classic PM. Focused on EMEA enterprise SaaS and consulting contexts.
1 PM vs Engagement Manager
Understanding the key shift in focus, scope and accountability.
| Dimension | Project Manager | Engagement Manager |
|---|
| Focus | Delivery execution | Business outcomes and client success |
| Relationship | Project team | C-suite and executive sponsors |
| Scope | Single project | Multiple workstreams or full programme |
| Commercial | Budget control | Revenue, upsell, contract management |
| Communication | Status updates | Strategic narrative and value articulation |
| Risk | Project risk | Commercial, reputational and strategic risk |
| Pre-sales | Rarely involved | Core part of the role — scope and estimate |
Key insight: An EM is not just a senior PM. The role requires commercial ownership, executive relationship management and the ability to grow the account — not just deliver the project.
2 Core EM Competencies
The six capabilities that define a strong Engagement Manager.
Executive Presence
Engage confidently with C-level stakeholders. Adapt communication style. Build trust quickly in high-pressure situations.
Value Articulation
Translate technical delivery into business outcomes — ROI, efficiency gains, risk reduction. Speak the client's language.
Commercial Acumen
Understand SOW, contract terms, billing milestones. Know when to raise a change request. Protect margin.
Escalation Management
Own difficult situations. De-escalate under pressure. Know when to involve leadership without losing client trust.
Pre-Sales Support
Work alongside sales to scope engagements, estimate effort and build customer confidence during the sales cycle.
Account Growth
Identify expansion opportunities organically during delivery. Turn delivery success into commercial opportunity.
3 Client Relationship Management
From project contact to trusted advisor.
Trusted Advisor Model
- Move beyond transactional PM: Clients should see you as a strategic partner, not just a delivery manager
- Proactive communication: Share bad news early, with a plan. Never let the client find out first
- QBR (Quarterly Business Review): Structured review of value delivered, roadmap and relationship health
- Success Criteria definition: Agree upfront how success will be measured — before work begins
- Executive Sponsor management: Regular touchpoints with senior client leadership separate from operational meetings
CSAT and NPS
CSAT
Customer Satisfaction Score. Typically 1–5 or 1–10. Measured per milestone or delivery phase.
NPS
Net Promoter Score. 0–10 scale. 9–10 = Promoters, 7–8 = Passives, 0–6 = Detractors.
Health Score
Composite metric combining CSAT, adoption, support tickets and engagement activity.
4 SaaS Implementation Delivery
The standard lifecycle for enterprise SaaS deployments.
🔍
DiscoveryUnderstand requirements and current state
📐
DesignSolution architecture and blueprinting
⚙️
ConfigureBuild and configure the solution
🧪
Test (UAT)Client validation against acceptance criteria
🎓
TrainEnd-user enablement and change management
🚀
Go-LiveCutover, hypercare and stabilisation
Critical Go-Live Elements
- Go-live criteria: Define acceptance criteria, rollback plan and hypercare period in advance
- Data migration: Extract, Transform, Load (ETL). Validate data quality before migration. Never skip dry runs
- Hypercare: Elevated support period post go-live — typically 2–4 weeks with dedicated resources
- Change management: Managing the human side of adoption. Training, communications, resistance management
5 Commercial Awareness
Contract types, SOW management and overrun prevention.
Contract Types
| Type | Description | Risk |
|---|
| Fixed Price (FP) | Agreed price regardless of actual cost | Low for client, high for vendor |
| Time & Materials (T&M) | Client pays actual hours + materials | Low for vendor, high for client |
| Fixed Price + Incentives | Fixed base with bonus for performance | Shared risk model |
| Cost Plus | Actual costs + agreed profit margin | Common in government / large programmes |
Managing Overrun — Lessons from Practice
- Track weekly, not monthly: Budget overruns compound quickly. Catch them early
- Scope creep is the #1 cause: Enforce change control from day one, without exception
- Undocumented requests: Every verbal agreement must be confirmed in writing. "I'll just do it quickly" is how projects overrun
- Time tracking tools: Enable early detection of effort vs estimate divergence
- Requirements matrix: Links each requirement to its source, design, test and delivery confirmation
- Lessons learned: Capture what drove overrun in retrospectives — prevent recurrence on the next programme
Key rule: Read the SOW before every change conversation. Any work outside the SOW requires a formal Change Request. Acceptance criteria define when payment milestones are triggered.
6 Stakeholder & Communication Management
Mapping, engaging and communicating with complex stakeholder landscapes.
Power / Interest Grid
High Power · High Interest
Manage closely. Regular updates, involve in key decisions, address concerns proactively.
High Power · Low Interest
Keep satisfied. Brief updates at milestones. Avoid overwhelming with detail.
Low Power · High Interest
Keep informed. They are often the day-to-day users — their adoption matters.
Low Power · Low Interest
Monitor. Minimal engagement unless their status changes.
Executive Communication
- Focus on impact, not detail: Summarise in 3 bullets — status, issues, next actions
- Channels formula: n(n-1)/2 — with 10 stakeholders there are 45 communication channels
- RAID log: Risks, Assumptions, Issues, Dependencies — review in every status meeting
- Escalation protocol: Define clear escalation paths before they are needed, not during a crisis
7 Leadership & Soft Skills
The human skills that determine EM success.
| Leadership Style | When to use |
|---|
| Directive | Crisis situations, new team members, tight deadlines |
| Coaching | Developing team capability, motivating high performers |
| Facilitative | Collaborative decisions, experienced cross-functional teams |
| Servant Leadership | Agile environments, removing blockers, empowering the team |
Negotiation: BATNA
BATNA (Best Alternative to a Negotiated Agreement): Know your walk-away point before entering any negotiation. The first number stated sets the reference point (anchoring) — use it deliberately. Separate people from the problem: negotiate on merits, not emotions.
Knowledge Track
AI Architect
Foundations of AI, machine learning, large language models, agentic systems and enterprise AI architecture. A practical reference for professionals who need to understand, govern and lead AI-powered transformation programmes.
1 AI Foundations
Core concepts every AI-adjacent professional must understand.
Key Definitions
| Term | Definition |
|---|
| Artificial Intelligence (AI) | Systems that simulate human intelligence — reasoning, learning, problem-solving, perception |
| Machine Learning (ML) | Subset of AI where systems learn from data without being explicitly programmed |
| Deep Learning | ML using multi-layer neural networks. Powers most modern AI applications |
| LLM | Large Language Model. AI trained on massive text datasets to understand and generate language (GPT, Claude, Gemini) |
| Foundation Model | A large, pre-trained model that can be fine-tuned for multiple downstream tasks |
| Generative AI | AI that creates new content (text, images, code, audio) based on patterns learned from training data |
| Inference | The process of running a trained model to generate predictions or outputs |
| Training | The process of teaching a model by exposing it to data and adjusting its parameters |
| Fine-tuning | Further training a pre-trained model on domain-specific data to specialise its capabilities |
2 Types of Machine Learning
The three paradigms and when each is used.
Supervised Learning
Learns from labelled data (input → output pairs). Used for classification and regression. Example: spam detection, credit scoring.
Unsupervised Learning
Finds patterns in unlabelled data. Used for clustering, anomaly detection. Example: customer segmentation.
Reinforcement Learning
Agent learns by interacting with an environment and receiving rewards/penalties. Used in robotics, game-playing, recommendation systems.
Semi-supervised
Combines small labelled dataset with large unlabelled dataset. Common when labelling is expensive.
3 Large Language Models (LLMs)
How they work, key concepts and enterprise applications.
How LLMs Work
- Transformer architecture: The underlying neural network design powering most modern LLMs (introduced in "Attention is All You Need", 2017)
- Tokens: LLMs process text as tokens (roughly 3/4 of a word). Context window = max tokens the model can process at once
- Temperature: Controls randomness of output. Low (0.1) = deterministic, High (0.9) = creative and varied
- Parameters: The learned weights of the model. GPT-4 has ~1.8 trillion; more parameters ≠ always better
- RLHF: Reinforcement Learning from Human Feedback — used to align models with human preferences and safety guidelines
Key LLM Concepts for Enterprise
| Concept | Description |
|---|
| Prompt Engineering | Crafting inputs to elicit optimal outputs. Includes system prompts, few-shot examples and chain-of-thought instructions |
| RAG (Retrieval Augmented Generation) | Augments LLM responses with real-time retrieval from a knowledge base. Reduces hallucinations, keeps answers current |
| Fine-tuning | Training a base model on proprietary data to specialise it for a specific domain or task |
| Hallucination | When an LLM generates confident but factually incorrect information. A key risk to manage in enterprise deployments |
| Context Window | Maximum amount of text the model can process in a single call. Ranges from 4K to 1M+ tokens |
| Embeddings | Numerical vector representations of text. Used for semantic search, similarity matching and RAG systems |
| Vector Database | Stores embeddings for fast similarity search. Examples: Pinecone, Weaviate, pgvector |
4 Agentic AI & Orchestration
The next evolution — AI systems that plan, act and use tools autonomously.
Agentic AI refers to AI systems that can autonomously perform multi-step tasks, use tools, call APIs, browse the web, write and execute code, and coordinate with other agents — with minimal human intervention at each step.
AI Agent
An LLM combined with tools and memory that can take actions in the world. Plans a sequence of steps to achieve a goal.
Tool Use
Agents call external tools: search, code execution, APIs, databases, file systems. The LLM decides when and how to use each tool.
Multi-Agent
Multiple specialised agents coordinating to complete complex tasks. One orchestrator delegates to specialist sub-agents.
Memory
Short-term (context window), long-term (vector DB) and episodic (conversation history). Enables agents to maintain state.
Enterprise Agentic Platforms
| Platform | Description |
|---|
| Salesforce Agentforce | Autonomous AI agents embedded in Salesforce CRM. Sales, service and marketing agents |
| ServiceNow AI Agents | Agents that handle IT workflows, approvals and service desk tasks autonomously |
| SAP Joule | SAP's generative AI assistant embedded across the SAP portfolio |
| Microsoft Copilot | AI agents integrated into Microsoft 365, Dynamics and Azure |
| LangChain / LangGraph | Open-source frameworks for building custom LLM applications and agent workflows |
| AutoGen / CrewAI | Multi-agent orchestration frameworks for complex autonomous workflows |
5 Enterprise AI Architecture
How AI systems are designed and integrated in enterprise environments.
Common Architecture Patterns
| Pattern | Description | Use Case |
|---|
| RAG Pipeline | Retrieve relevant documents → augment prompt → generate response | Internal knowledge base Q&A, support bots |
| LLM + API | LLM calls external APIs to retrieve or act on real-time data | Booking, CRM updates, live data retrieval |
| Fine-tuned Model | Base model trained on proprietary domain data | Legal, medical, financial document analysis |
| Orchestration Layer | Middleware that routes queries to appropriate models/agents | Complex enterprise AI products |
| Semantic Cache | Cache LLM responses based on semantic similarity to avoid redundant API calls | Cost reduction at scale |
Key Infrastructure Components
- LLM Provider: OpenAI (GPT-4), Anthropic (Claude), Google (Gemini), Meta (Llama), Mistral
- Vector Database: Pinecone, Weaviate, Qdrant, pgvector (PostgreSQL extension)
- Orchestration Framework: LangChain, LangGraph, AutoGen, CrewAI, Semantic Kernel
- Embedding Model: Converts text to vectors. OpenAI text-embedding-3, Cohere, sentence-transformers
- Observability: LangSmith, Helicone, Arize — monitoring LLM calls, latency and costs
- Cloud AI Services: Azure OpenAI, AWS Bedrock, Google Vertex AI — enterprise-grade LLM hosting
6 AI Governance & Ethics
Responsible AI deployment in enterprise contexts.
EU AI Act
EU regulation (2024) classifying AI systems by risk level. High-risk systems require documentation, human oversight and conformity assessment.
Trustworthy AI
Principles: Lawful, Ethical, Robust. Includes transparency, fairness, privacy, accountability (EU AI HLEG framework).
AI Risk
Hallucination, bias, data leakage, prompt injection, model drift, over-reliance. Each requires specific mitigation.
Human-in-the-Loop
Keeping humans as decision points for high-stakes AI outputs. Critical in healthcare, legal, financial and HR applications.
Data Privacy in AI
- GDPR: Personal data must not be sent to third-party LLMs without appropriate safeguards. Use on-premise or private cloud models for sensitive data
- Data residency: Ensure AI infrastructure complies with data sovereignty requirements
- PII handling: Implement data masking or anonymisation before sending to external models
- Model cards: Documentation describing a model's training data, intended use, performance and limitations
7 AI Applied to PM & Engagement Management
How AI is changing the PM and EM role in practice.
The shift: AI does not replace the PM or EM. It eliminates the administrative burden, surfaces risks earlier and accelerates reporting — freeing the human to focus on judgement, relationships and strategy.
| PM/EM Task | AI Application |
|---|
| Status reporting | AI drafts weekly reports from project data, Jira tickets and meeting notes |
| Risk identification | ML models analyse historical project data to flag early warning indicators |
| Meeting summaries | AI transcribes, summarises and extracts action items from calls |
| Document analysis | LLMs review SOWs, contracts and requirements documents to extract key obligations |
| Resource planning | AI optimises resource allocation across projects based on skills, availability and priority |
| Client communication | AI drafts escalation emails, status updates and executive summaries |
| Knowledge management | RAG systems make lessons learned and best practices searchable and actionable |
Key Tools for AI-Augmented Delivery
- n8n / Zapier: Workflow automation connecting AI models with project management tools
- Notion AI / Confluence AI: AI-assisted documentation and knowledge management
- GitHub Copilot / Cursor: AI code generation for technical PMs and solution architects
- Otter.ai / Fireflies: AI meeting transcription and action item extraction
- Anthropic Claude / OpenAI GPT-4: General-purpose LLMs for drafting, analysis and research
8 AI Glossary
Key terms every AI-aware professional should know.
| Term | Definition |
|---|
| Attention Mechanism | Core component of Transformers allowing the model to weigh the importance of different tokens |
| Chain-of-Thought | Prompting technique asking the model to reason step-by-step before answering |
| Embedding | Numerical vector representing text, images or data in a high-dimensional space |
| Grounding | Connecting AI outputs to verified real-world information to reduce hallucinations |
| Hallucination | AI generating confident but factually incorrect information |
| MCP | Model Context Protocol — standard for connecting AI models to external tools and data sources |
| Multimodal | AI that processes multiple input types: text, images, audio, video |
| Prompt Injection | Attack where malicious instructions in input override system instructions |
| RAG | Retrieval Augmented Generation — retrieving relevant context before generating a response |
| System Prompt | Instructions given to an LLM before the user conversation, defining its behaviour and constraints |
| Token | Smallest unit of text processed by an LLM. Roughly 3/4 of an English word |
| Tool Use / Function Calling | Ability of an LLM to call external APIs or functions to retrieve or act on information |
| Vector Database | Database optimised for storing and searching embeddings by semantic similarity |
| Zero-shot | Asking an LLM to perform a task with no examples. Few-shot = providing 1–5 examples |