CPM AI Phase 1: Business Understanding - Practice Flashcards

Flashcards covering business understanding, AI patterns, DIKUW pyramid, ROI considerations, PoC vs pilot, go/no-go framework, trustworthy AI, team roles, and real-world examples from CPM AI.

What is the first question to answer before starting an AI project according to CPM AI?

Identify the objective (the general business or organizational objective) and determine if it requires a cognitive AI solution.

What are the seven AI patterns CPM AI uses to frame problems?

Recognition, Conversation, Pattern/Anomaly Detection, Predictive Analytics, Autonomous Systems, Goal-Driven Systems, Hyper Personalization.

What does the DIKUW pyramid stand for?

Data, Information, Knowledge, Understanding, Wisdom.

Why is it often better to pursue a pilot rather than a Proof of Concept in CPM AI?

Pilots use real-world data and production environments; PoCs are lab-based and may not translate to real-world adoption or ROI. MVP/Pilot aids iterative learning.

What is the AI go/no-go assessment and its three traffic-light categories?

A nine-light framework assessing Business Feasibility, Data Feasibility, and Implementation Feasibility; greens mean go, yellows indicate risk, reds stop.

What is the difference between augmented/assisted intelligence and autonomous systems?

Augmented/assisted AI supports humans and typically offers faster ROI; autonomous AI operates with little to no human involvement and usually has longer ROI and higher risk.

What factors influence whether a project should be automated or AI-powered?

Cost, speed of implementation, deployment complexity, data quality, labor costs, risk tolerance, and need to learn from data.

What makes a good AI ML project in CPM AI terms?

Solves a short-term organizational need; is cognitively appropriate; scalable; leverages the DIKUW sweet spot; and may be staged as smaller projects.

Where does machine learning sit in the DIKUW pyramid?

Knowledge—the 'how' of deriving insights from data and identifying patterns.

Which AI patterns are harder to shortcut with foundation models and why?

Hyper Personalization and Goal-Driven/Autonomous systems often require custom, data-specific solutions beyond generic models.

What are typical ROI considerations when selecting AI projects?

Short-term ROI, revenue growth/differentiation, cost reduction, improved operations, scalability; avoid projects with long, uncertain ROI or misalignment.

What is the role of MVP in CPM AI methodology?

Minimal Viable Product used in pilots to validate ROI and feasibility in real-world settings before broader rollout.

What are the three major question groups in AI go/no-go assessment?

Business feasibility (problem definition, change willingness, ROI), Data feasibility (data availability/quality/quantity), Implementation feasibility (technology skills, timing, usability).

What is the significance of trustworthy AI requirements in phase one?

Identify bias/fairness, privacy, safety, transparency, consent, governance, and responsible AI considerations to avoid blockers later.

Why is interpretability and explainability important in CPM AI phase one?

Regulatory, legal, and stakeholder needs require that AI decisions can be understood and traced; reduces risk and builds trust.

What are the four main areas to form an AI project team?

Business (line of business, analysts), Data Science (domain experts, data scientists), Data Engineering (data/cloud engineers), Operationalization (app/devs, admins).

What data issues are examined in data feasibility?

Whether the required data exists, its quantity/access, and its quality for training and measuring success.

Where does the term 'citizen data scientist' fit in CPM AI?

A non-traditional data scientist who builds ML outputs using no-code/low-code tools rather than formal training.

What is the purpose of the DIKUW pyramid beyond data collection?

To guide where AI adds value—from turning data into information and knowledge, toward understanding and wisdom, focusing on patterns and insights.

What types of data are central to NASA’s predictive maintenance example?

Time-series sensor data (e.g., vibration, emissions, turbine readings) used to detect anomalies and predict remaining useful life.

What does Coca-Cola’s brand content moderation example illustrate about AI patterns?

Recognition pattern; objective is to detect brand content and NSFW content; discusses false positives/negatives and the trade-offs with human moderation.

Why are real-world pilots emphasized over PoCs in AI projects?

Pilots use real data and production environments, providing a truer test of ROI, usability, and deployment challenges.

How should you approach pattern selection relative to ROI in CPM AI?

Identify the business objective or ROI target first, then select the AI pattern(s) that best achieve that objective rather than forcing a pattern to fit the ROI.

What is meant by ‘think big, start small, iterate often’ in CPM AI?

Break projects into manageable iterations with quick time-to-ROI, validating and expanding scope as you learn.

What is the role of ‘off-the-shelf’ tools and foundation models in CPM AI?

They can shortcut many AI patterns (conversational, recognition, summarization, etc.), enabling faster ROI and leaving custom development to later stages.