# Nisha's problem (/claude-and-chatgpt-2/case-1-nishas-ai-team/nishas-problem) --- type: Document title: "Nisha's problem" description: "Who Nisha is, why she wants to learn US GAAP, why she decides to write a book, and how she sets up a team of AI Workers to help her." status: stable order: 0.51 ksor: owner: team:panaversity audience: [ public ] approval: by: process:panaversity at: 2026-11-09T20:42:36Z chapter: "F00C" part: Front expert_status: required concepts: [ "Nisha's problem" ] last_verified: 2026-11-09 sources: - id: ifrs-roadmap title: "IFRS Foundation, Jurisdictional Roadmap Development Tool (March 2025)" resource: https://www.ifrs.org/content/dam/ifrs/around-the-world/roadmap-development-tool/jurisdictional-roadmapdevelopmenttool-2025.pdf - id: practical-log title: "Nisha's AI Team: Practical Log recording the exercise of November 8, 2026" resource: "appendix-the-practical-log.md" generated: at: 2026-11-09T20:42:36Z by: process:panaversity trust_tier: unverified build_id: sha256:46fd53c5d7ab1c14f6347908d45c9fdaecda4d914439923ee878bdcecd57803e dirty: true ksor_version: 0.0.60 --- > **Recorded on November 8, 2026.** This page tells what Nisha actually did that evening, as the [practical log](appendix-the-practical-log.md) records it. The [lab](lab-take-nishas-method-to-2-0.md) later asks you to try newer features that Nisha did not use. ## A Sunday evening On a Sunday evening in November 2026, Nisha opened her laptop. She had two browser tabs open: one for Claude and one for ChatGPT. Next to them was an empty Word file. She planned to use it as a diary of everything she did that night. Nisha was a chartered accountant, and she had won a gold medal in her exams. She was very good at her job. But tonight she was not checking anyone's accounts. Tonight she wanted to plan a book, and she wanted two AI tools to do most of the writing. One question was on her mind. Could two AI subscriptions do the work of a small team? And if they could, what would her own job be? ## Nisha's problem Nisha was a member of the Institute of Chartered Accountants of Pakistan (ICAP), the country's best-known accounting body. Chartered accountants (CAs) prepare and check company accounts. These accounts are published so that banks, lenders and investors can see how a company is doing. CAs also give advice on tax and on running a business. Nisha earned Rs 10 lakh (one million rupees) a month. That is a very good salary in Pakistan. But Nisha liked international brands. She wore them, and she ate at expensive places. Her rent alone was Rs 5 lakh (500,000 rupees) a month. After tax and her other bills, she was spending more than she earned. There was a second worry. People kept telling her that AI would soon replace accountants. One day, someone said to her: *"You are working for the Pakistani market. You will never earn much here. Why not work for foreign clients? Try freelancing, or get a remote job."* ## Two languages of accounting Nisha liked the idea. She set her sights on the "Big 4", the four largest accounting firms in the world: Deloitte, EY, KPMG and PwC. All four have large practices in the United States. Then she found a problem. Every country needs rules for how to prepare accounts. Each rule document is called a *standard*, and it usually covers one topic, such as inventory. Most of the world, including Pakistan, uses one set of rules called **IFRS** (International Financial Reporting Standards). More than 140 countries and regions require it, mainly for listed companies (companies whose shares are traded on a stock market).[^ifrs-roadmap] The United States uses a different set, called **US GAAP** (US Generally Accepted Accounting Principles). The two sets are a bit like British and American English. Most of it is the same, but some rules are different, and some words are used differently. For example, the standard accounting word for goods waiting to be sold is "inventory" (IFRS says "inventories"). The everyday word "stock" can mean those goods or shares in a company, depending on the context. British accounts have often used "stock" for goods, while American accounts usually keep "stock" for shares. Here is one real difference. US GAAP allows a method called LIFO ("last in, first out") for working out the cost of the goods a company has sold. Under LIFO, the cost of the most recently bought goods is counted first, even if older goods actually left the warehouse first. IFRS does not allow LIFO. People often say that IFRS is *principle-based* (it gives general ideas and asks you to use judgment) while US GAAP is *rule-based* (it gives detailed rules). Nisha knew IFRS very well. She knew much less about US GAAP. So she had two problems: 1. She needed to learn US GAAP properly. 2. Even after learning it, she had to prove it. Why would a US employer believe that an accountant from Pakistan was an expert in US rules? ## The idea: a book written with AI Workers A mentor gave Nisha an idea: *write a book*. A book comparing IFRS and US GAAP would help other accountants make the same move. Writing it would force her to learn US GAAP. And the finished book would show employers what she knew. Writing a book alone takes a long time, so Nisha decided to build a team of AI Workers. She bought two subscriptions, Claude and ChatGPT, for about Rs 30,000 a month each: about Rs 60,000 a month in total. (Like her salary, these prices are made up for the case.) Nisha liked to say she had hired "ten AI workers" for the price of two subscriptions. She meant the many jobs she could give the tools, such as writer and reviewer, not that they would do the work of ten people. Both tools had recently been upgraded. This book calls the upgraded apps Claude 2.0 and ChatGPT 2.0. These are not official product names. The tools could now search the web, read sources and work on a document in several rounds. In the scenario, this was the big change: Claude could improve a draft, grade it and improve it again by itself, where before a user had to ask for each round by hand. In the practical, Claude did run three such rounds in a single reply ([Step 3](steps-1-to-3-claude-writes-and-grades-its-own-work.md#step-3-claude-improves-its-own-work)). Nisha used the upgraded apps in their simplest form: one project in each app, with prompts she typed and replies she read. She did not use separate threads, scheduled tasks or ChatGPT's Work toggle. Nisha gave each worker a job: - **Claude** would be the main writer. - **ChatGPT** would be the second reviewer, a fresh pair of eyes. - **Nisha** would be the manager and the final checker. She also wrote a short brief. Her readers would be accountants trained in IFRS who want to work with US GAAP, and the other way round. To keep things simple, she would start with only two parts of the book: the *preface* (the short introduction at the start of a book) and the *table of contents*. Nisha followed a six-step method ([Exhibit 1](exhibits.md#exhibit-1-nishas-method-six-steps-with-human-checks-in-between)). Claude writes. Claude grades its own work. Claude improves the work by itself, aiming for 9.5 out of 10 in each area. ChatGPT reviews it. Claude improves it again using ChatGPT's review. Finally, a person checks the facts and decides when the work is ready. Nisha added one rule of her own: **she would also review the work herself between the steps.**[^practical-log] Next, [Steps 1 to 3: Claude writes and grades its own work](steps-1-to-3-claude-writes-and-grades-its-own-work.md). [^ifrs-roadmap]: The IFRS Foundation's Jurisdictional Roadmap Development Tool (March 2025) states that more than 140 jurisdictions already require the use of IFRS Accounting Standards. Each jurisdiction decides which companies must use them. [^practical-log]: Nisha's AI Team: Practical Log, section 2 (Method).