PMI Certified Professional in Managing AI

Most AI solutions fall into a small number of patterns. You learn to recognize recognition, conversation, prediction, anomaly detection, personalization, autonomy and goal-driven optimization, and match each to the problems it solves.

• Recognition: images, speech and text
• Conversation and human interaction
• Prediction and decision support
• Patterns and anomaly detection
• Personalization
• Autonomous systems and goal-driven optimization
• Combining patterns in one solution

You build A pattern map that matches five business problems to the AI patterns that suit them.

Trust is designed in, not added at launch. You set ethical principles, identify bias, privacy and transparency risks, put governance in place, and track the laws and standards that apply.

• Ethical principles for AI projects
• Sources of bias and ways to test for it
• Privacy, consent and data rights
• Transparency and explainability
• AI governance, accountability and oversight
• Regulation and standards that affect the project
• Recording risks and decisions in a risk log

You build A responsible AI plan with risks, controls and accountable owners.

Many AI projects fail because the problem was the wrong one. You test business needs for value, check whether AI is the right tool, judge feasibility, and write the case a sponsor can approve.

• Turning a business need into an AI use case
• Is AI the right solution? Rules, analytics and AI compared
• Value, cost and return estimates
• Feasibility: data, skills, technology and time
• Build, buy or adapt a foundation model
• Defining success criteria and measures
• Getting sponsor and stakeholder agreement

You build A business case and feasibility summary for one AI use case.

Data decides what a model can learn. You list the data a solution needs, find where it lives, check rights and quality, and decide whether it is ready or what has to change first.

• Data requirements from the use case
• Finding internal and external sources
• Data quality dimensions and profiling
• Rights, privacy and licensing of data
• Labeling needs and annotation effort
• Assessing data readiness and gaps

You build A data requirements and readiness report for one AI use case.

Preparation takes much of the effort on an AI project. You plan cleaning, integration, feature work and splits for training and testing, and guard against bias and leakage as you go.

• Data pipelines and integration
• Cleaning, handling missing values and outliers
• Feature selection and engineering
• Training, validation and test sets
• Labeling quality and managing annotators
• Detecting bias and leakage in prepared data
• Versioning and documenting datasets

You build A data preparation plan with a dataset specification and quality checks.

A project manager does not need to build the model, but must steer the work. You choose approaches with the team, manage training and experiments, handle foundation models and prompts, and keep development tied to the business need.

• Choosing a modeling approach with the team
• Supervised, unsupervised and reinforcement learning at a glance
• Generative models, fine-tuning and prompt-based solutions
• Planning experiments and tracking them
• Managing compute, cost and timelines
• Handling uncertainty in model progress
• Documentation and reproducibility

You build A model development plan with experiments, milestones and decision points.

A good score is not the same as a useful model. You choose metrics that reflect business value, set thresholds, analyze errors, test for fairness and reliability, and decide whether to iterate, adjust the scope or stop.

• Accuracy, precision, recall and other metrics
• Linking metrics to business outcomes
• Setting acceptance thresholds
• Error analysis and edge cases
• Fairness, safety and reliability testing
• Evaluating generative output
• Deciding to iterate, release or stop

You build An evaluation report that links model metrics to business success criteria.

A model in a notebook helps no one. You plan deployment, integration and infrastructure, prepare people and processes for change, and set up the operating model that keeps the solution working.

• Deployment options: cloud, edge and embedded
• Integrating a model with business systems
• MLOps and release practices
• Change management and user adoption
• Service levels, support and ownership
• Security of models and pipelines

You build A deployment and adoption plan for one AI solution.

Models decay as the world changes. You monitor performance, data and fairness, detect drift, schedule retraining, handle incidents, and decide when a solution should be retired.

• Monitoring performance and data quality
• Model and data drift
• Retraining triggers and schedules
• Feedback loops and human review
• AI incidents and response
• Auditing and reporting to governance
• Retiring a model

You build A monitoring and maintenance plan with drift thresholds and retraining triggers.

Phases overlap and loop back in an AI project. You plan iterations, coordinate business, data and technical teams, manage risk and communications, and keep sponsors informed when results are uncertain.

• Iterative project phases and the loops between them
• Coordinating business, data and technical roles
• Estimating when outcomes are uncertain
• Risk and issue management on AI projects
• Stakeholder communication and expectations
• Closing an AI project and handing it to operations

You build An AI project plan with phases, roles, risks and a communications approach.

One initiative carries the whole course. You select the use case, check data and responsibility issues, plan development and evaluation, and set out how the solution will be deployed and monitored.

• Choosing and justifying the use case
• Assessing data and responsible AI risk
• Planning development and evaluation
• Planning deployment and monitoring
• Presenting the plan to a sponsor

You build A complete charter and delivery plan for an AI initiative.