Google Cloud Generative AI Leader Certification Exam Prep
Master all four sections of the current Generative AI Leader exam guide at business level, with no coding, from first principles to confident scenario "best answer" decisions across Google Cloud's gen AI products. Independent of Google and no pass is promised.
Read lesson 1 free →Try 10 practice questions free →
Levels 1 and 2 are free to read, no account needed. Levels 3 and 4 are free with a free account, and so are 20 answers from the AI tutor. 10 questions from the bank behind this course's timed mock are free to try, no account needed.

Expert · 42 levels · 4 free · Created Oct 2026 · Professionally curated by levelupwith.com
What's inside
- Level 1: The Exam and What Gen AI IsFree
[Fundamentals of gen AI] Sets out the exam format (50-60 multiple-choice questions in 90 minutes, four weighted sections), states that this course is independent of Google and promises no pass, then places generative AI inside AI, machine learning and deep learning.- State the exam format and the four sections with their approximate weightings (30%, 35%, 20%, 15%)
- Explain how AI, machine learning, deep learning and generative AI nest inside one another
- Tell a generative AI task (producing new content) from a predictive ML task (classifying or forecasting) in a business scenario
- Define natural language processing (NLP) in plain business language and explain how it relates to AI, machine learning, large language models and generative AI.
- Level 2: Three ML ApproachesFree
[Fundamentals of gen AI] Teaches supervised, unsupervised and reinforcement learning from first principles and what business problem each one fits.- Define supervised, unsupervised and reinforcement learning in plain business language
- Match a scenario (predicting from past examples, finding hidden groups, learning by reward) to the right approach
- Rule out the two wrong approaches in a 'best answer' question and say why
- Level 3: Foundation Models, LLMs and MultimodalFree account
[Fundamentals of gen AI] Explains what a foundation model is, how a large language model is one kind of foundation model, and what makes a model multimodal.- Define foundation model, large language model (LLM) and multimodal model, and state how they relate
- Explain why one foundation model can be adapted to many tasks, unlike a traditional single-task ML model
- Identify from a scenario whether an LLM or a multimodal model is needed
- Level 4: Diffusion Models and Prompt TuningFree account
[Fundamentals of gen AI] Completes the core concepts with diffusion models for image and video generation and the exam's contrast between prompt engineering and prompt tuning.- Describe at business level how a diffusion model generates images by refining noise
- Distinguish prompt engineering (wording the input by hand) from prompt tuning (adapting how the model responds to prompts through training)
- Pick the right term when a scenario describes one of these concepts without naming it
- Level 5: ML Lifecycle Stages and Their Tools
[Fundamentals of gen AI] Walks through the ML lifecycle stages the exam guide names (data ingestion, data preparation, model training, model deployment, model management) and the Google Cloud tool associated with each.- List the ML lifecycle stages in order and say what happens in each
- Match each stage to a Google Cloud tool that supports it
- Identify which stage a described business activity belongs to
- Level 6: Choosing a Foundation Model
[Fundamentals of gen AI] Teaches the factors used to select a foundation model for a business use case: modality, context window, security, availability and reliability, cost, performance, fine-tuning and customization.- Name the selection factors for a foundation model and explain each in one sentence
- Weigh trade-offs such as cost against performance, or context window against speed, for a given scenario
- Choose the most suitable model profile when a scenario states its requirements
- Level 7: Create, Summarize, Discover, Automate
[Fundamentals of gen AI] Introduces the four categories of gen AI business use case (create, summarize, discover and automate) across text, image, code, audio and video.- Define the create, summarize, discover and automate use-case categories
- Classify a described business use case into the correct category
- Suggest a realistic gen AI use case in each category for a given department
- Recognise data analysis and personalized user experience as gen AI business use cases, alongside text, image, code and video generation, and place each in the create, summarize, discover or automate c
- Level 8: Data Quality for AI
[Fundamentals of gen AI] Explains why data quality and accessibility determine AI results, using the dimensions the exam names: completeness, consistency, relevance, availability, cost and format.- Define each data quality and accessibility dimension and its effect on model output
- Diagnose which data quality problem is causing poor results in a scenario
- Explain to a stakeholder why a strong model cannot make up for poor data
- Level 9: Data Types: Structured and Labeled
[Fundamentals of gen AI] Distinguishes structured from unstructured data and labeled from unlabeled data, and links each to the business uses and ML approaches it supports.- Classify example business data as structured or unstructured
- Explain the difference between labeled and unlabeled data and which ML approaches need which
- Identify the data type a scenario involves and what it implies for the solution
- Level 10: The Five Layers of Gen AI
[Fundamentals of gen AI] Maps the gen AI landscape into its five layers (infrastructure, models, platforms, agents and gen AI-powered applications) and what each contributes to a business solution.- Name the five layers in order and describe the role of each
- Place a described product or capability in the correct layer
- Explain how the layers depend on one another when a business adopts gen AI
- Level 11: Gemini and Gemma
[Fundamentals of gen AI] Introduces Google's Gemini family of multimodal foundation models and the Gemma family of lightweight open models, and contrasts when each is the better fit.- State what Gemini is and what kinds of input and output it is used for
- State what Gemma is and why an organization would pick an open, lightweight model
- Choose between Gemini and Gemma for a scenario based on its control, size and deployment needs
- Level 12: Imagen and Veo
[Fundamentals of gen AI] Covers Imagen for image generation and Veo for video generation, then practises choosing between all four Google foundation models.- State what Imagen is for and what Veo is for
- Link Imagen and Veo to the diffusion concept from Level 4 at business level
- Select Gemini, Gemma, Imagen or Veo as the best answer for a described business need
- Level 13: Google Cloud's Three Strengths
[Google Cloud's gen AI offerings] Explains the strengths the exam attributes to Google Cloud in gen AI: an AI-first approach, an enterprise-ready platform and an open ecosystem.- Explain what AI-first, enterprise-ready and open each mean in the exam guide's terms
- Match a customer concern, such as innovation, security and reliability, or avoiding lock-in, to the strength that answers it
- Recognise which strength a scenario statement is describing
- Level 14: AI Hypercomputer, TPUs and GPUs
[Google Cloud's gen AI offerings] Covers the infrastructure under Google Cloud's gen AI: the AI Hypercomputer as an integrated system, and TPUs and GPUs as the accelerators that train and run models.- Describe the AI Hypercomputer as Google Cloud's integrated AI infrastructure and what it is for
- Contrast TPUs (Google's custom AI chips) with GPUs (general-purpose accelerators) at business level
- Identify these as the infrastructure layer and explain why it matters to a business leader
- Level 15: Keeping Control of Your Data
[Google Cloud's gen AI offerings] Explains how Google Cloud lets an organization keep control of its data when using gen AI, through security, privacy, governance and reliability commitments.- Explain what 'data control' means for an organization using Google Cloud gen AI
- Describe at business level the security, privacy and governance assurances an enterprise looks for
- Choose the answer that addresses a stakeholder's data-control concern in a scenario
- Level 16: Low-Code Paths to Gen AI
[Google Cloud's gen AI offerings] Shows how Google Cloud's low-code and no-code options let business teams use and build gen AI without programming, and when a prebuilt product beats a custom build.- Explain what low-code and no-code mean and why they widen access to gen AI
- Distinguish a ready-made product, a low-code build and a custom-coded build by effort, skill and flexibility
- Recommend the least complex option that meets a scenario's requirements
- Level 17: Gemini App, Gemini Advanced and Gems
[Google Cloud's gen AI offerings] Covers the Gemini app as Google's AI assistant, Gemini Advanced as the paid tier named in the exam guide, and Gems as customized versions of Gemini for repeated tasks.- State what the Gemini app, Gemini Advanced and Gems each are and what each is for
- Explain how a Gem saves a user from repeating the same instructions
- Pick among the three for an individual-productivity scenario
- Level 18: Gemini Enterprise
[Google Cloud's gen AI offerings] Introduces Gemini Enterprise as the organization-wide product for searching company data and using agents, with the Gemini Notebook API, multimodal search and custom agents, noting any older name Google's documentation still uses.- Describe what Gemini Enterprise is for and who in an organization uses it
- Explain the purpose of the Gemini Notebook API, multimodal search and custom agents within it
- Contrast Gemini Enterprise with the Gemini app in a 'best answer' scenario
- Level 19: Gemini for Google Workspace
[Google Cloud's gen AI offerings] Explains Gemini for Google Workspace as gen AI built into tools such as Gmail, Docs, Sheets, Slides and Meet, and contrasts it with the Gemini app and Gemini Enterprise.- Describe how Gemini for Google Workspace helps inside everyday productivity apps
- Give a create, summarize, discover or automate example for a Workspace app
- Decide between Gemini for Google Workspace, the Gemini app and Gemini Enterprise for a scenario
- Level 20: Agent Search vs Google Search
[Google Cloud's gen AI offerings] Contrasts Agent Search (older name Vertex AI Search), which builds search over an organization's own data, with Google Search over public web information.- State what Agent Search (older name Vertex AI Search) is for and what data it searches
- Explain how it differs from Google Search in data source and business purpose
- Choose the right one when a scenario needs answers from private company content or from the public web
- Level 21: Customer Engagement Suite: Four Products, Four Jobs
[Google Cloud's gen AI offerings] Learn what the Customer Engagement Suite is for and tell apart its four parts: Conversational Agents (self-service virtual agents), Agent Assist (real-time help for human agents), Conversational Insights (analysis of customer conversations) and Contact Center as a Service (CCaaS, the cloud contact centre itself).- State in one line what Conversational Agents, Agent Assist, Conversational Insights and CCaaS are each for
- Pick the right Customer Engagement Suite product for a contact-centre scenario, such as deflecting routine calls versus coaching a live human agent
- Explain why customer-facing virtual agents and human-agent assistance are different business needs, even though both use gen AI
- Level 22: Agent Platform: Model Garden and AutoML
[Google Cloud's gen AI offerings] Learn what Agent Platform (older name Vertex AI) is as Google Cloud's unified platform for building with AI, and how Model Garden (a catalogue for discovering and choosing models) differs from AutoML (training a custom model on your own data without writing code).- Describe Agent Platform as the builder platform and recognise it under its older name, Vertex AI
- Explain what Model Garden is for and the kinds of models it offers: Google, partner and open models
- Decide in a scenario whether a team should choose an existing model from Model Garden or train a custom one with AutoML
- Level 23: Agent Platform: RAG APIs and Custom Agents
[Google Cloud's gen AI offerings] Learn how Agent Platform lets a business connect models to its own data through RAG APIs and build custom agents that reason, use tools and complete multi-step tasks.- Explain at business level what Agent Platform's RAG APIs are for: answering from an organisation's own documents
- Describe what makes something an agent rather than a chatbot: a model, a goal, tools and a reasoning loop
- Contrast building a custom agent on Agent Platform with using a ready-made product such as Gemini Enterprise or the Customer Engagement Suite
- Level 24: Agent Tools: Extensions, Functions, Data Stores, Plugins
[Google Cloud's gen AI offerings] Learn the four kinds of agent tooling the exam guide names (extensions, functions, data stores and plugins) and what each lets an agent do beyond generating text.- Define extensions, functions, data stores and plugins and say what each is for
- Match a business need (calling an external API, running the company's own code, looking up company content) to the right tool type
- Explain why tools are what let an agent act on real-time information and outside systems
- Level 25: Cloud Services Behind Agents: Storage, Databases, Run
[Google Cloud's gen AI offerings] Learn the supporting Google Cloud services an agent relies on: Cloud Storage for files and unstructured data, databases for structured records, Cloud Functions for small event-driven pieces of code, and Cloud Run for running containerised applications without managing servers.- State what Cloud Storage, databases, Cloud Functions and Cloud Run each contribute to a gen AI solution
- Choose between Cloud Storage and a database for a given type of data
- Contrast Cloud Functions (a single event-triggered function) with Cloud Run (a whole containerised service) in a scenario
- Level 26: The Pre-Built AI APIs: Speech, Language, Documents, Vision
[Google Cloud's gen AI offerings] Learn the eight pre-trained Google Cloud AI APIs the exam guide names and the single job each does: Speech-to-Text, Text-to-Speech, Translation, Document Translation, Document AI, Vision, Video Intelligence and Natural Language.- Name the input and output of each of the eight APIs
- Separate the confusable pairs: Speech-to-Text vs Text-to-Speech, Translation vs Document Translation, Document AI vs Vision, Vision vs Video Intelligence
- Recognise when a pre-built API meets a need more simply than a custom model or a foundation model
- Level 27: Agent Studio vs Google AI Studio
[Google Cloud's gen AI offerings] Learn when to use Agent Studio (older name Vertex AI Studio), the enterprise prompt-and-build workspace inside Agent Platform on Google Cloud, and when to use Google AI Studio, the lightweight tool for quickly trying Gemini models with an API key.- Describe what each studio is for and who typically uses it
- Choose the right studio in a scenario by weighing enterprise governance, security and data control against fast prototyping
- Place both studios correctly alongside the rest of section 2: Gemini app, Gemini Enterprise, Agent Platform and the pre-built APIs
- Level 28: Where Foundation Models Fall Short
[Techniques to improve gen AI model output] Learn the built-in limits of foundation models that the later techniques exist to fix: dependence on training data, the knowledge cutoff, bias, hallucinations and poor handling of edge cases.- Define knowledge cutoff, hallucination, data dependency, bias and edge case in plain language
- Identify which limit is at work in a described model failure
- Explain why these limits follow from how a model is trained, rather than being faults a better prompt alone removes
- Level 29: Grounding: First-Party, Third-Party and World Data
[Techniques to improve gen AI model output] Learn what grounding is (tying a model's answers to verifiable sources) and the three kinds of source: first-party enterprise data, third-party data and world data, such as grounding with Google Search.- Define grounding and explain how it reduces hallucinations and works around the knowledge cutoff
- Distinguish first-party, third-party and world data with an example of each
- Choose the right grounding source for a scenario, such as internal policy questions versus current public events
- Level 30: Retrieval-Augmented Generation (RAG)
[Techniques to improve gen AI model output] Learn how RAG works step by step: retrieve relevant content, add it to the prompt, then generate the answer. Learn how Google Cloud offers this through Agent Search (older name Vertex AI Search) and Agent Platform's RAG APIs.- Describe the retrieve, augment and generate steps without technical jargon
- Explain how RAG relates to grounding: RAG is a common way of achieving it
- Contrast the ready-made Agent Search with building retrieval yourself using RAG APIs
- Level 31: Fine-Tuning and Choosing the Right Fix
[Techniques to improve gen AI model output] Learn what fine-tuning is (further training a foundation model on your own examples to specialise its behaviour) and how to choose between prompt engineering, grounding or RAG, and fine-tuning by cost, effort and the problem at hand.- Explain what fine-tuning changes in a model, and what it does not: it does not supply fresh facts
- Compare prompting, RAG and fine-tuning by cost, time, data needed and the limit each addresses
- Select the least costly technique that solves a described output problem
- Level 32: Keeping Output Reliable: HITL, Monitoring and Drift
[Techniques to improve gen AI model output] Learn how output quality is protected after launch: human in the loop (HITL) review for high-stakes decisions, continuous monitoring of model performance, detecting drift as real-world data changes, and Feature Store for managing and serving consistent model features.- Recognise scenarios where HITL is the right safeguard and say what the human reviewer adds
- Define drift and explain why a model that worked at launch can degrade over time
- State what monitoring and Feature Store are each for within Agent Platform (older name Vertex AI)
- Explain what model versioning is for and why keeping track of model versions supports reliable, repeatable gen AI output after launch.
- Explain how automatic upgrades, security patches and updates on Google-managed models and APIs keep a gen AI solution current and reduce the maintenance work left to the business.
- Level 33: Prompting by Example: Zero-, One-, Few-Shot and Role
[Techniques to improve gen AI model output] Learn the foundational prompting techniques: zero-shot (no examples), one-shot (one example), few-shot (several examples) and role prompting (assigning the model a persona or point of view).- Classify a given prompt as zero-shot, one-shot, few-shot or role prompting
- Explain when adding examples improves output and when zero-shot is enough
- Combine a role with examples to fix a prompt that gives output in the wrong tone or format
- Level 34: Advanced Prompting: Chaining, Chain-of-Thought, ReAct
[Techniques to improve gen AI model output] Learn three techniques for complex tasks: prompt chaining (splitting work into a sequence of prompts), chain-of-thought (asking the model to reason step by step) and ReAct (reason and act, alternating reasoning with tool use).- Define prompt chaining, chain-of-thought and ReAct, and tell them apart in a scenario
- Explain why ReAct is the pattern behind agents that use tools
- Choose the technique that fits a multi-step, reasoning-heavy or action-taking task
- Level 35: Sampling Parameters and Safety Settings
[Techniques to improve gen AI model output] Learn how model settings shape output without changing the prompt: sampling parameters such as temperature, top-p and output length (token count) control randomness and size, while safety settings filter harmful content.- Explain what raising or lowering temperature and top-p does to creativity versus consistency
- Recommend settings for a scenario, such as factual support answers versus brainstorming
- Describe what safety settings do and how they differ from sampling parameters
- Level 36: Choosing the Right Gen AI Solution for a Need
[Business strategies for a successful gen AI solution] Learn how to start from the business need rather than the technology, weighing requirements, available data, skills, cost, time and risk to decide between a ready-made product, a low-code build and a custom solution.- Translate a business problem into the requirements a gen AI solution must meet
- Compare buy, configure and build options using the Google Cloud offerings from section 2
- Identify the factors that most often decide the best answer in a scenario: data, skills, budget, time to value and risk
- Level 37: Integrating Gen AI Into the Organisation
[Business strategies for a successful gen AI solution] Learn the steps for bringing a gen AI solution into an organisation: connecting it to existing systems and workflows, aligning stakeholders, preparing people for change and scaling from pilot to production.- Outline the steps from identifying a use case through pilot to scaled rollout
- Explain why stakeholder buy-in, training and change management matter as much as the technology
- Spot integration risks in a scenario, such as siloed data or a solution with no clear owner
- Level 38: Measuring the Impact of a Gen AI Initiative
[Business strategies for a successful gen AI solution] Learn how to judge whether a gen AI solution is working by defining success up front and tracking business measures such as return on investment, productivity, cost, quality, adoption and customer satisfaction.- Choose key performance indicators that match the stated goal of a gen AI project
- Distinguish business impact measures from technical model quality measures
- Use measurement results to decide whether to scale, adjust or stop an initiative
- Level 39: Secure AI: SAIF, IAM and Security Command Center
[Business strategies for a successful gen AI solution] Learn why security must cover the whole ML lifecycle and what each named item is for: Google's Secure AI Framework (SAIF) as the guiding framework, Identity and Access Management (IAM) for controlling who can do what, and Security Command Center for visibility into risks and threats.- Describe SAIF as a framework for securing AI systems, not a product to switch on
- Contrast IAM (access control) with Security Command Center (security posture and threat visibility)
- Pick the right security measure for a scenario at a given stage of the ML lifecycle
- Describe what Google Cloud's secure-by-design infrastructure contributes to protecting a gen AI solution and tell it apart from IAM and Security Command Center.
- State what workload monitoring tools are for in securing a running gen AI solution and choose them over IAM or Security Command Center when a scenario calls for watching workloads in operation.
- Level 40: Responsible AI: Transparency and Privacy
[Business strategies for a successful gen AI solution] Learn the first half of responsible AI: being transparent about how and when AI is used, protecting personal data, and the difference between anonymization (identity removed irreversibly) and pseudonymization (identifiers replaced, so re-identification is possible with a key).- Explain why transparency builds trust and what it requires of a business deploying gen AI
- Distinguish anonymization from pseudonymization and choose the right one for a scenario
- Identify privacy risks in a described gen AI use case and a suitable mitigation
- Level 41: Responsible AI: Bias, Fairness, Accountability, Explainability
[Business strategies for a successful gen AI solution] Completes responsible AI by teaching, at business level, how bias enters a gen AI system and what fairness requires, who is accountable for a model's outputs, and why explainability matters, each contrasted with the transparency and privacy ideas already covered so the four terms can be told apart in a scenario.- Identify the responsible AI principle at stake in a business scenario by distinguishing bias, fairness, accountability and explainability from one another and from transparency and privacy
- Explain where bias can enter a gen AI solution (training data, prompts, human review) and choose the most appropriate business-level step to reduce unfair outcomes
- Select the best answer in scenario-based practice questions about who is accountable for a gen AI system's decisions and when an explainable output is needed for trust or compliance
- Level 42: Timed Mock Exam: 60 Questions in 90 Minutes Timed mock
[All sections] A timed mock exam of 60 scenario-based multiple-choice practice questions in 90 minutes with a 70% pass mark, testing all four exam guide sections in proportion to their weightings (fundamentals ~30%, Google Cloud's offerings ~35%, techniques to improve output ~20%, business strategies ~15%); these are practice questions written for this course, which is independent of Google, not real exam questions, and passing the mock does not guarantee a pass on the real exam.
Free 10 questions from the bank behind this course's timed mock are free to try, no account needed. Read all 10 with their answers →
Access
Levels 1 and 2 are free to read, no account needed. Levels 3 and 4 are free with a free account, and so are 20 answers from the AI tutor. 10 questions from the bank behind this course's timed mock are free to try, no account needed. Every level, the podcast edition and the AI tutor come with All Access at £4.99/month or any Creator plan — see pricing.
More like this
- Google Cloud Associate Cloud Engineer Exam Prep
Covers every item in Google's Associate Cloud Engineer exam guide, with each section given levels in line with its exam weight.
Exam prep · Google · Cloud & Compute · Expert · 52 levels · 4 free · Professionally curated by levelupwith.com
- Support Pro: Helping Users Move to Google Workspace
Get first and second line support ready to carry users through a Microsoft 365 to Google Workspace cutover.
Course · Google · Suite Support Pro: Configure and Fix · Expert · 40 levels · 4 free · Professionally curated by levelupwith.com
- Google Workspace Support Pro: Gemini for End Users
Diagnose and fix real Gemini tickets in Google Workspace.
Course · Google · Suite Support Pro: Configure and Fix · Expert · 40 levels · 4 free · Professionally curated by levelupwith.com
- Google Workspace Support Pro: Drive and Shared Drives
Diagnose and fix Google Drive problems the way experienced support engineers do: read Admin console Drive policy and explain what it does, find out who can…
Course · Google · Suite Support Pro: Configure and Fix · Expert · 34 levels · 4 free · Professionally curated by levelupwith.com
- Google Workspace Support Pro: Meet and Chat
Diagnose and fix the Meet and Chat faults that reach first and second line support: audio, video, screen sharing, spaces, external guests, recordings and room…
Course · Google · Suite Support Pro: Configure and Fix · Expert · 40 levels · 4 free · Professionally curated by levelupwith.com
- Google Workspace Support Pro: Gmail and Calendar
Configure delegation, send-as, shared inboxes, calendars and resources.
Course · Google · Suite Support Pro: Configure and Fix · Expert · 40 levels · 4 free · Professionally curated by levelupwith.com