Claude and ChatGPT 2.0
- Approved by
- Panaversity ·
Work better with AI every day, and build the foundation to do this work for clients.
Welcome to The AI Agent Factory. Until now, most people asked an AI a question and read its answer. Now you can hand it a whole job. This book teaches you to do that well.
An AI Worker is an AI system you hand a whole job to, inside limits you set, the way you would hand a job to a skilled coworker. It does the work, and you check what comes back. Many people call this an AI agent, and the title of this book does too. Inside the book, we use a more exact name, AI Worker. The Preface explains the difference.
Here is an example of a whole job: "Read these 40 invoices, find the ones that do not match the bank statement, and write me a one-page summary." A limit might be: the AI Worker may read the files, but it may not send any email.
This first section of the book is called Claude and ChatGPT 2.0. The "2.0" is this book's nickname for that change, from asking questions to handing over whole jobs. It is not an official product name.1
We start with the new Claude and ChatGPT desktop apps, the programs you install on your computer. For both, this book uses the desktop app, not the website or the phone app. We put this section first for a reason: we want you using these apps on real work from your first day. The later parts of the book explain how and why AI Workers work, and teach you to build your own.
What this section gives you
This section gives you two things.
- First, for yourself. You learn to use AI well in your everyday work, as a knowledge worker: a person who works with information, such as an accountant, an analyst or a manager.
- Then, for clients. You build the foundation to set up this way of working for other people, as a Forward Deployed Engineer (FDE). An FDE goes into a company, finds a real business problem, and builds an AI Worker to solve it, inside clear limits.
By the end, you can hand an AI Worker a whole job, set its limits, and check what comes back.
It is one foundation with two uses. The skill is the same. What changes is who the work is for, and how much a mistake costs.
How you learn it. This section ends with four cases. A case is the story of one person doing a real job: a goal, some limits, and a decision to make. You decide what you would do. Then you try it in a lab, a guided exercise you do yourself, and see whether your decision works. The other pages in this section teach you the features. The cases teach you judgment: deciding well when no rule gives you the answer. So do not skip them.
All four cases follow one person, Nisha. She is an accountant, not an FDE. But in each case her job gets bigger and a mistake costs more. That is the road toward FDE work.

What you need. A Mac or Windows computer, and free Claude and ChatGPT accounts. You must be 18 or older, in a country where the apps are offered. You need no programming. The programming comes later in the book. Some of the labs work best on a paid plan. Those labs also give you steps for a free account.
Why now. Anthropic makes Claude. OpenAI makes ChatGPT. Both companies are now moving into this kind of client work.
- Anthropic has released ten ready-made templates for AI agents for finance jobs. One of them helps prepare a company's accounts at the end of each month, the same kind of job you will meet in Case 4.2
- Anthropic and OpenAI have each started a company, paid for by large investors, whose engineers work with client companies to set up AI for them.3 OpenAI calls its engineers Forward Deployed Engineers.4
- Finance is not the only field. Both companies have also built tools for hospitals and other care providers,5 for medical research,6 and for finding security problems in software and suggesting fixes.7
The same kind of work is reaching many fields. What would it look like in yours?
How this section teaches
Business schools have taught through cases for more than a hundred years.8 They use cases to train judgment, because judgment grows with practice. Later, Chapter 0.4. Critical Thinking and Judgment in the AI Era gives you a method for checking AI work. You practice first, here. Critical thinking gives you the method, and the cases give you the practice. Together they build judgment, the skill that matters most when you apply AI.
Why cases fit this book so well
Cases are a good way to teach many subjects. For AI Workers, they fit very well, for six reasons.
-
The skill is judgment, not buttons. When you work with an AI Worker, the buttons are easy, and the pages before the cases teach them. The hard part is deciding: which job to hand over, what limits to set, when to stop the work, and how to check what comes back. A case puts you in these decisions again and again, until checking becomes a habit.
-
Managing AI Workers is like managing people. You hand a job to a coworker and you check the result. Many managers learn this through cases. Managing an AI Worker is the same kind of skill, so it can be learned the same way.
-
The apps change, but the decisions stay. The Claude and ChatGPT apps change every few months, so the pages on features are updated often. The decisions in a case last longer: what to trust, what to check, what stays your responsibility. The feature pages show you how. The cases teach you when and why.
-
Cases give you experience before you have it. Many readers have never handed a whole job to an AI, and some are new to work itself. A case lets you follow someone else's job, including the mistakes, before you face your own.
-
You can test your answer. A business school case ends with a discussion, because no one can run a real company's decision again. Here you can. You build the AI Worker in the lab, run it, and see whether your decision works. A classroom discussion cannot do that.
-
It is the same work a Forward Deployed Engineer does. An FDE works inside a company to solve its business problems with AI Workers. Every company is a new case: a real problem, real limits, people with different goals, and a decision about what to build. Later in this book, you learn to do this work. The cases prepare you for it, because an FDE's job is a case that never ends.
By Case 4, you will have practiced, at small scale, the steps at the center of FDE work: understand a real problem, set limits, build the worker, and check its work. These skills are yours even if you never want the job title.
In this section
Read these in order. Two of them are optional. For a map of the whole book, see How to Use This Book.
- The New Way of Working: what has changed, one job from start to finish, and what stays your responsibility. It ends with a practice lab you can do on a free account.
- Tour of the Apps: each feature of both apps, side by side. It is optional. Skim it now, or come back when a page uses a feature.
- Behind the Screen: what happens while an AI Worker does a job, and what keeps the job safe.
- The Architecture: how the apps are built, the evidence behind them, and what a business must add. It is optional and more technical.
- Build Something: how to make documents, slides, sheets and other finished pieces of work in both apps. Do it before the cases.
- Case 1: Nisha's AI Team: a case study of an accountant who manages two AI tools to plan a book in one evening, with a lab that takes her method further.
- Case 2: Nisha's First Client: a case study of her first paid job, run with an AI Worker that works overnight, with a lab where you build the same worker.
- Case 3: Nisha's Firm: a case study of her firm of three, with one AI Worker in the firm's team chat that all of them can ask, and a lab where you test who can make it send a file to a client.
- Case 4: Nisha's Month-End: a case study of a monthly report that took two days by hand. The firm joins the features it already uses into one setup, with a lab where you build it, and the final assessment for this whole section.
Check yourself
Question 1 / 5
0 answered
In this book, what is an AI Worker?
Sources
- The name comes from a video walkthrough of the new Claude desktop app, which quotes one of the engineers who worked on Claude Code calling it "basically a Claude 2.0" (https://www.youtube.com/watch?v=tM0dwXhOv2o). Applying it to both apps is this book's own choice.↑
- Anthropic says: "We're releasing ten ready-to-run agent templates for the most time-consuming work in financial services." Each one "ships as a plugin in Claude Cowork and Claude Code, and as a cookbook for Claude Managed Agents". One is a "Month-end closer", which "runs the close checklist, prepares journal entries, and produces close reports". Agents for financial services, May 5, 2026 (https://www.anthropic.com/news/finance-agents).↑
- Anthropic says the new company "will work with mid-sized companies across sectors" and that "Applied AI engineers from Anthropic will work alongside the firm's engineering team". Building a new enterprise AI services company with Blackstone, Hellman & Friedman, and Goldman Sachs, May 4, 2026 (https://www.anthropic.com/news/enterprise-ai-services-company).↑
- OpenAI says the company will "embed engineers specialized in frontier AI deployment, known as Forward Deployed Engineers, or FDEs, into organizations". OpenAI launches the OpenAI Deployment Company, May 11, 2026 (https://openai.com/index/openai-launches-the-deployment-company/).↑
- OpenAI announced OpenAI for Healthcare, for hospitals and other care providers, on January 8, 2026 (https://openai.com/index/openai-for-healthcare/). Anthropic announced Claude for Healthcare a few days later (https://www.fiercehealthcare.com/ai-and-machine-learning/jpm26-anthropic-launches-claude-healthcare-targeting-health-systems-payers).↑
- Anthropic announced Claude for Life Sciences on October 20, 2025 (https://www.anthropic.com/news/claude-for-life-sciences). OpenAI announced GPT-Rosalind, a model for biology and drug research, on April 16, 2026 (https://openai.com/index/introducing-gpt-rosalind/).↑
- Anthropic introduced Claude Code Security in February 2026, and OpenAI introduced Codex Security on March 6, 2026. Both look for weak points in software and suggest fixes. OpenAI unveils Codex Security to automate code security reviews, Axios, March 6, 2026 (https://www.axios.com/2026/03/06/openai-codex-security-ai-cyber).↑
- Harvard Business School began teaching with cases in the early 1920s. Its MBA students now read more than 500 cases in their two years (https://poetsandquants.com/2022/04/13/a-century-of-cases-harvard-marks-100th-birthday-of-the-case-method/).↑
- Owner
- Panaversity