<aside> <img src="/icons/link_pink.svg" alt="/icons/link_pink.svg" width="40px" />

Watch meeting recording: Click here

</aside>

Today’s topics

1. What’s New: GPT-5 Agent Mode
2. OpenAI Playground & Prompt Optimization
3. Practical examples on how to use Agent mode

1. Agent Mode

Agent Mode makes GPT-5 behave less like a chatbot and more like a thinking assistant.

Instead of giving a quick answer and moving on, it works through a problem step by step, remembers what it has already done, and explains how it reached the final conclusion.

<aside> 📌

Think of it as the difference between asking someone on the street for advice versus hiring a consultant. A normal model gives you a quick suggestion. Agent Mode lays out the process: gather information, check it, compare options, and then recommend a path forward.

</aside>

What Agent Mode Does

• Breaks problems into steps → handles complex questions in order, instead of giving a shallow answer.
• Keeps track of context → remembers earlier steps, decisions, and any constraints you set (budget, KPIs, timelines).
• Follows a workflow → collect data → validate it → reason about it → recommend an outcome.

Example

If you ask: “What should our pricing strategy be?”

• Normal AI: generic suggestion.

• Agent Mode:

  1. Collect competitor prices
  2. Check accuracy of the data
  3. Compare with your profit margins
  4. Consider customer behavior
  5. Recommend pricing with reasons behind it

  <aside> 📌

  You can connect your apps to your ChatGPT so Agent can access it and work with MCP connections

  </aside>