Build AI Workflows and AI Agents with OpenAI Agent Builder

Stay updated for 2026 and ready to submit feedback via the next lecture form to request resources, keeping the course content accurate in a collaborative learning journey.

Build a deep research agent with the OpenAI agent builder, following a topic to keywords to researched final report flow, exploring templates, guardrails, JSON outputs, and while loops.

Explore OpenAI agent templates and guardrails, focusing on the core logic, set state, and tools; learn jailbreak guardrails, PII moderation, prompt injection detection, and intent classification for robust agent workflows.

Design an AI agent that infers a topic from user input using set state and guardrails, and outputs a topic research JSON schema, for building workflows with OpenAI Agent Builder.

Verify your organization in settings to enable verification for access to models like reasoning and image generation, using a cost-free, country-supported ID not used in the last 90 days.

Set up a dynamic keyword inference workflow that outputs a json array of keywords with a default of four, and include a user approval step to approve or reject.

Set state and transform keywords to build AI workflows with an OpenAI agent, handling keyword extraction, transform operations, and user approval to drive deep research.

Explain while loops for non-programmers by illustrating how a loop runs while a condition is true, avoids infinite loops, and iterates from zero to four across keywords.

Learn to conduct deep research by looping through keywords with a while loop, tracking current research and current keyword, and applying a deep research agent to each keyword.

Debug by aligning keyword data with the keyword agent, validating the state and set state steps, and finalizing a rephrasing agent to produce digestible research outputs.

Learn to build a goal-driven deep research agent that plans, acts through tools and APIs, maintains memory, and uses user-approved guardrails for looped tasks.

Discusses desired enhancements for the OpenAI Agent Builder, contrasting workflows with autonomous AI agents, and notes cycles, complex inputs, and the need for simpler forms.

Set the AI's system message to define its role, tone, scope, and guardrails, and learn how prompt structure, templates, and hierarchies shape performance, outcomes, and guard against prompt leakage.

Explore jailbreaking in AI workflows and agents, examining how system messages, north stars, rules, and roles, along with instruction safeguards, protect secrets and deter abuse in LLM-based tools.

Engage in a Harry Potter-themed character guessing game with six levels of hints revealed through chat with AI, while exploring how to protect secrets in AI apps and workflows.

Share feedback via the next lecture's form to turbocharge the course and sustain high energy while refining AI workflows and agents with OpenAI Agent Builder.

Design and build an AI assistant using RAG and MCP to support AI workflows, embed data, and connect everywhere, enabling web browsing, calendar access, and email checks.

RAG retrieves up-to-date information, augments prompts with that content, and generates grounded answers, keeping data fresh and sources visible across domains from healthcare to finance.

Set up an AI assistant for your startup by configuring startup knowledge, a tailored system message, and vector store integration, and enable tools like web search for crisp insights.

Vector stores act as a memory bank for AI, turning data into embeddings and meaning vectors, enabling search by meaning rather than exact words.

Upload text-based documents to the vector store to build a knowledge base, supporting Word, PDFs, and PowerPoint, and test the AI agent for multilingual real-time hospitality scenarios.

Introduce MCP, an open standard that connects AI models to data sources through a single host–client–server flow. Use primitives—prompts, resources, tools, routes, and sampling—to build universal, secure connections.

See how MCP in action automates invoicing with stripe, using a SQL-driven data fetch and approval loop. MCP is the USB-C plug for AI, enabling safe tool access through protocol.

Learn how to set up an MCP server with Zapier, obtain and configure the OpenAI API key, and connect Gmail, tasks, and calendar tools to automate workflows.

Finalize the AI assistant by diagnosing a web search issue, reconnecting Zapier MCP and web browsing, and coordinating Google calendar and tasks with API keys.

Diogo Alves de Rezende shares his data-driven background, a management and analytics MSc from Esmt Berlin, and how data guides restaurant pricing and menu optimization.

Harness no-code tools to build your own website with an OpenAI agent, connect it to a lovable-built site, and deploy a ready prototype that answers queries on your LinkedIn profile.

Build AI workflows and AI agents with OpenAI Agent Builder to deliver your own website as the key output, while the chat kit chatbot cannot be included.

Create a personal website with lovable by drafting prompts from your LinkedIn, outlining three sections: hero, professional, and projects and interests, and leveraging introductory credits.

Draft a system prompt to build a chatbot knowledgeable about you, set voice and style, keep verbosity low, and publish the agent using the chat kit workflow.

Set up a ChatKit chatbot by configuring security domains, generating keys and workflow IDs, and integrating the OpenAI SDK, with practical steps and documentation guidance.

Attempt to build the AI agent and integrate a chatbot into a website, troubleshooting failed integrations while exploring workflow automation with AI tools and limited credits.

Design ai driven product and marketing workflows with the OpenAI agent builder, connecting managers to multiple agents using if-else logic and widgets for on demand company knowledge and brainstorming.

Learn to build AI workflows with the agent builder by routing the chief of staff's outputs to product and marketing via if-else conditions for delegation.

Explore AI workflows and agents for marketing and product teams, from chief of staff routing tasks to product and marketing managers, to defining features and user stories for restaurants.

Create and publish a simple widget with ChatKit using the agent builder, mapping text and images to an output of top features, reasoning, and user stories, then preview and publish.

Shows planning ux improvements for the menu management page within the agent builder, demonstrates a screenshot workflow and an image-attachment error, and notes ongoing course updates.