GenAI Leader cheat sheet

Fundamentals of gen AI

Generative AI Concepts: Foundation Models, LLMs, and Business Use Cases

• Generative AI is a branch of machine learning that creates new content; traditional ML classifies or predicts.
• A foundation model is pretrained on broad data and adapted to many tasks; an LLM is a foundation model specialized in language.
• Multimodal foundation models like Gemini handle text, images, audio, and video in a single model.
• Diffusion models generate images by learning to reverse noise — the concept behind Imagen; Veo generates video.
• Supervised learning uses labeled data, unsupervised finds patterns in unlabeled data, reinforcement learns from rewards and penalties.
• The ML lifecycle runs data ingestion, data preparation, model training, model deployment, and model management — as a loop.
• Choose a foundation model by modality, context window, security, availability, cost, performance, fine-tuning, and customization — matched to the scenario.
• Gen AI business value maps to four verbs: create, summarize, discover, and automate.

Data Types in Gen AI: Quality, Structured vs Unstructured, Labeled Data

• Gen AI output quality is bounded by data quality: incomplete, inconsistent, or irrelevant data produces confident but wrong answers.
• Data quality and accessibility have six exam characteristics: completeness, consistency, relevance, availability, cost, and format.
• Structured data fits predefined rows and columns (transactions, CRM records, inventory tables); unstructured data is free-form (emails, contracts, images, audio, video).
• Unstructured data is the large majority of enterprise data, and foundation models are the first technology to unlock it at scale.
• Labeled data carries correct-answer tags and powers supervised learning; unlabeled data has none and powers unsupervised learning and foundation-model pretraining.
• Labeling is expensive human work — a key reason adapting a pretrained foundation model beats building a labeled dataset from scratch.
• Poor data quality hits the business as lost trust, bias, inflated cost, misleading decisions, and missed use cases.
• Assess data quality and accessibility before scaling a gen AI initiative, not after deployment.

The Five Layers of the Gen AI Landscape and What They Mean for Business

• The gen AI landscape has five layers: infrastructure, models, platforms, agents, and applications, and each builds on the layer below.
• Infrastructure (TPUs, GPUs, AI Hypercomputer, Google Cloud data centers) should be consumed from a cloud provider, not built.
• The models layer is a choice among foundation models like Gemini, Gemma, Imagen, and Veo, not a mandate to train your own.
• Platforms such as Agent Platform with Model Garden are where enterprises build custom, data-grounded solutions with control over security and governance.
• Agents combine a model with reasoning and tools to complete multi-step tasks, which makes them powerful for automation but demands stronger governance.
• Applications like the Gemini app, Gemini for Google Workspace, and Gemini Enterprise deliver the fastest time to value with no development team.
• Lower layers trade speed for control and differentiation; higher layers trade customization for speed and low cost.
• Most organizations operate at two layers: applications for broad productivity, plus platforms or agents for a few differentiating use cases.

Google's Foundation Models: Gemini, Gemma, Imagen, and Veo Explained

• Google's four named foundation models are Gemini (multimodal reasoning), Gemma (lightweight open models), Imagen (image generation), and Veo (video generation).
• Gemini is natively multimodal with a long context window, making it the choice for reasoning, conversation, document analysis, code, and mixed-input tasks.
• Gemma's open weights let organizations run and fine-tune the model on their own infrastructure, trading peak capability for control, customization, and cost efficiency.
• Imagen generates and edits still images from natural-language descriptions; Veo generates high-definition video from text or image prompts.
• Choose by modality first (image work to Imagen, video work to Veo, reasoning and language to Gemini or Gemma), then by control needs (open and self-hosted points to Gemma).
• Standard selection factors still apply: context window, cost, performance, security, availability and reliability, and the need for fine-tuning or customization.
• The models are complements: real projects combine them, and all four are accessible to builders through Model Garden on Agent Platform.

Google Cloud's gen AI offerings

Google Cloud's Gen AI Strengths: Platform, Infrastructure, and Openness

• Google's AI-first approach and research heritage — including the Transformer architecture — create a short path from breakthrough to product, so customers get state-of-the-art capability and a strong roadmap.
• Enterprise-ready means five qualities: responsible, secure, private, reliable, and scalable — match the quality to the concern the exam scenario stresses.
• The comprehensive ecosystem integrates the same Gemini models across the Gemini app, Google Workspace, Gemini Enterprise, and Google Cloud, so skills, governance, and trust carry across every surface.
• The open approach — Gemma open models, Model Garden's first-party, open, and third-party catalog, and open frameworks — preserves choice and reduces lock-in.
• AI-optimized infrastructure combines the AI Hypercomputer architecture, custom-designed TPUs, the latest GPUs, and efficient global data centers, consumed on demand through cloud computing.
• Your data is not used to train Google's foundation models without permission; security, privacy, and governance controls keep enterprise data inside your boundary.
• Low-code and no-code tools, pre-trained models, and APIs democratize AI development so domain experts can build without ML engineering skills.

Prebuilt Gen AI at Work: Gemini App, Gemini Enterprise, and Workspace

• Prebuilt offerings are finished products — fastest time to value, no ML skills required — and the entry point of Google Cloud's gen AI spectrum.
• The Gemini app is a standalone multimodal assistant for individuals; Gemini Advanced adds the most capable models and Gems.
• Gems are custom, reusable versions of Gemini configured once with instructions — the answer when a scenario needs a repeatable personal specialist.
• Gemini Enterprise grounds AI in company data with multimodal search, custom agent capabilities, and the Gemini Notebook API, all permission-aware.
• Gemini for Google Workspace embeds AI in Gmail, Docs, Sheets, Slides, Meet, and Chat — assistance in the flow of work, on the user's own content.
• Distinguish the three by scope and grounding: general knowledge for one person (Gemini app), the user's own files (Workspace), the whole organization's data (Gemini Enterprise).
• Business value scales with scope: personal productivity, workforce-wide time savings, and unlocked institutional knowledge, respectively.

Improving Customer Experience with Agent Search and Customer Engagement Suite

• Google Cloud improves customer experience through two families: external search offerings (Agent Search, grounding with Google Search) and the Customer Engagement Suite.
• Agent Search, on the Gemini Enterprise Agent Platform, embeds Google-quality generative search — with prebuilt RAG and grounded, cited answers — into a company's own sites and apps.
• Agent Search grounds answers in your enterprise content; grounding with Google Search adds fresh world knowledge — know which grounding a scenario needs.
• Conversational Agents resolve customer requests directly via chat and voice, blending deterministic flows with generative answers grounded in company content.
• Agent Assist is a real-time copilot for human agents: suggested responses, surfaced knowledge, transcription, and automatic summaries.
• Conversational Insights analyzes all conversations at scale for topics, sentiment, and contact drivers — enabling root-cause fixes and evidence-based coaching.
• Contact Center as a Service is the cloud-native platform underneath: omnichannel routing and telephony, elastic scale, no on-premises infrastructure.
• Business benefits to cite: higher search conversion, contact deflection, lower handle time, faster agent ramp-up, and rising customer satisfaction.

How Google Cloud Empowers Developers: Agent Platform and RAG

• Agent Platform is Google Cloud's unified environment for discovering models, grounding them in company data, and building, deploying, and managing gen AI agents.
• Model Garden is a curated library of Google, open, and partner foundation models — its value is informed model choice under one set of enterprise controls.
• Agent Search delivers out-of-the-box, Google-quality search and grounded, source-referenced answers over an organization's own data.
• Agent Platform AutoML trains custom models from a company's labeled data through a guided workflow, with no training code or deep ML expertise required.
• Prebuilt RAG with Agent Search is the managed retrieval path — fastest time to value; RAG APIs are the component path for teams that need control over each pipeline step.
• Custom agents built on Agent Platform go beyond answering questions: they use instructions, grounding, and tools to complete multi-step business tasks.
• The selection rule: take the most packaged offering that meets the need — every step toward custom adds engineering cost that a specific requirement must justify.

Gen AI Agent Tooling: Extensions, Functions, Data Stores, and Plugins

• Tools are the agent's connection to the external environment — without them a model can only generate text from frozen training data; with them an agent can retrieve live information and complete tasks.
• Extensions are standardized bridges to external APIs, executed by the agent platform on the agent's behalf.
• Functions keep execution on your side: the model outputs which function to call and with what arguments, but your application validates and runs the action — the governed choice for consequential operations.
• Data stores ground an agent in company documents and data, delivering current, cited answers that stay accurate through content updates, with no retraining.
• Plugins extend an agent into outside applications and services, reusing capabilities the business already has.
• Cloud Storage backs files and documents, databases back live records, Cloud Functions runs lightweight action logic, Cloud Run hosts containerized services, and Agent Platform is where agents are built and governed.
• Pre-built AI APIs — Speech-to-Text, Text-to-Speech, Translation, Document AI, Cloud Vision, Video Intelligence, Natural Language — give agents ready-made abilities with speech, language, images, video, and documents.
• Google AI Studio is for quick prototyping with Gemini models; Agent Studio is for building grounded, governed, production gen AI agents.

Techniques to improve gen AI model output

Overcoming Foundation Model Limitations: Grounding, RAG, and HITL

• Foundation model limitations all stem from data dependency: the model only knows what its training data contained, up to its knowledge cutoff.
• Grounding connects model output to verifiable sources; RAG implements it by retrieving relevant enterprise content and inserting it into the prompt before generation.
• Hallucinations and stale answers point to grounding and RAG as the fix; generic tone or format points to prompt engineering; persistent domain gaps point to fine-tuning.
• Human in the loop (HITL) is the standing safeguard for high-stakes and fairness-sensitive decisions, and reviewer feedback improves the system over time.
• The cost-effective sequence is prompt engineering first, grounding and RAG next, fine-tuning last — escalate only when the cheaper technique falls short.
• Continuous monitoring covers KPIs, performance tracking, drift monitoring, versioning, automatic model upgrades, and security patches — launch is the start, not the end.
• The Agent Platform Feature Store centrally manages feature data so models and agents consume consistent, governed inputs across environments.

Prompt Engineering Techniques: Few-Shot, Role, and Chain-of-Thought

• Prompt engineering is deliberately designing the model's input — instruction, context, format, examples — and is the fastest, cheapest way to improve gen AI output.
• Zero-shot means no examples, one-shot means one, few-shot means several; add examples when output misses your format or standard, and use the fewest that work.
• Role prompting assigns a persona that steers tone, depth, and vocabulary — but it changes how the model communicates, not what it factually knows.
• Prompt chaining breaks complex work into linked steps with inspectable intermediate outputs, moving quality control into the middle of the process.
• Chain-of-thought prompting asks for step-by-step reasoning before the answer, improving multi-step logic and making conclusions auditable.
• ReAct prompting interleaves reasoning with actions such as search or tool lookups, and is the conceptual foundation of AI agents.
• Techniques combine in practice: a role plus few-shot examples inside a chain, with ReAct gathering live information, is a typical production pattern.

Grounding Techniques: RAG, Google Search, and Sampling Settings

• Grounding connects LLM responses to verifiable data sources, reducing hallucinations and keeping answers current without retraining the model.
• First-party data is what your organization owns, third-party data is acquired from external providers, and world data is broad public knowledge such as Google Search results.
• RAG retrieves relevant content at query time and generates from it, making output more accurate, current, and attributable with citations.
• Pre-built RAG with Agent Search is the fastest path to grounding on enterprise content; RAG APIs give developers component-level control over a custom retrieval pipeline.
• Grounding with Google Search grounds responses in fresh world data, directly addressing the model's knowledge cutoff for public, time-sensitive questions.
• Temperature controls randomness (low = focused and consistent, high = varied and creative); top-p limits token selection to the smallest set whose probabilities sum to p.
• Safety settings set configurable thresholds for blocking harmful content, and output length caps response size — both tuned per use case, not fixed.
• Tokens are the unit of gen AI cost and capacity: token count and output length are how leaders control spend and response size.

Business strategies for a successful gen AI solution

Implementing a Transformational Gen AI Solution: The Google Cloud Steps

• A transformational gen AI solution starts from a business need and works backward to the technology, never the reverse.
• The main solution types are text generation, image generation, code generation, and personalized user experiences, and each maps to distinct business problems.
• Business requirements (outcomes, users, budget, compliance) and technical constraints (data quality, integration, skills, security) jointly determine the right solution.
• Choose the simplest approach that meets the requirement; every layer of customization adds cost, maintenance, and risk.
• The recommended integration sequence is: define vision and use cases, assess readiness, select the solution, pilot, establish governance and security, integrate and scale, then measure and iterate.
• A pilot converts uncertainty into evidence cheaply, and a pilot that prevents a costly mistake is a success.
• Impact measurement requires KPIs defined before launch and a baseline to compare against; ROI weighs total value against total cost, including governance and training.
• Most gen AI failures are strategic, caused by skipping a recommended step, not by technical shortcomings of the models.

Secure AI and SAIF: Protecting AI Systems on Google Cloud

• Secure AI means protecting an AI system's data, models, infrastructure, and interactions from attacks and misuse across the entire ML lifecycle.
• Know the core threats conceptually: data poisoning corrupts training data, prompt injection manipulates a deployed model through its inputs, model theft steals the model asset, and evasion and leakage round out the list.
• Security must cover every lifecycle stage, data, training, deployment, and monitoring, because each stage exposes different assets and the weakest stage defines overall risk.
• SAIF is Google's conceptual framework for securing AI systems: it extends existing security foundations to AI, brings AI into threat detection, and keeps controls consistent, automated, and matched to business risk.
• SAIF's business benefits are reduced breach risk, faster and more confident AI adoption, stronger compliance posture, and preserved customer trust.
• Match the Google Cloud tools to their purpose: secure-by-design infrastructure (protected foundation with default encryption), IAM (least-privilege access control), Security Command Center (centralized posture and threat visibility), workload monitoring (anomaly detection on running systems).
• Security on Google Cloud is shared: Google secures the platform; the customer governs access, data, and acceptable use.

Responsible AI in Business: Transparency, Privacy, Bias, and Fairness

• Responsible AI is a business requirement, not just an ethical ideal — it protects trust, brand reputation, and regulatory standing.
• Google's AI Principles commit to socially beneficial AI that avoids unfair bias, is safe, accountable to people, and privacy-first — but the deploying business always retains accountability for its own use.
• Transparency means disclosing when and how AI is used; concealment turns ordinary AI limitations into trust-destroying incidents.
• Anonymization irreversibly removes identifying information; pseudonymization replaces identifiers with reversible tokens kept separately — pseudonymized data is still personal data under most privacy laws.
• Model outputs reflect training data: poor or unrepresentative data produces biased, unfair outcomes at scale, so data quality is a fairness issue.
• Accountability requires named owners, governance processes, and human oversight — especially a human in the loop for high-stakes decisions.
• Explainability, including grounding answers in cited sources, turns opaque AI assertions into verifiable ones and is often a regulatory expectation for consequential decisions.