# A company of one (/claude-and-chatgpt-2/case-1-nishas-ai-team/a-company-of-one) --- type: Document title: "A company of one" description: "What Nisha's diary showed at the end of the evening, and the new question it raised about her job." status: stable order: 0.54 ksor: owner: team:panaversity audience: [ public ] approval: by: process:panaversity at: 2026-11-09T20:42:36Z chapter: "F00C" part: Front expert_status: required concepts: [ "Managing AI Workers" ] last_verified: 2026-11-09 sources: - 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. ## What Nisha's diary showed Before going to sleep, Nisha read her diary from the start. She wrote down what she had seen, without deciding yet what it meant: 1. **Size.** She asked for a description and an outline, and the first draft was 2,470 words. She cut it to 1,531 words. By the end it had grown back to 2,599 words, and the sources had grown from 4 to 15. 2. **Scores.** Claude gave its own version 3 a 9.5 for accuracy and quality. ChatGPT gave the same version 8 and 7. Claude then gave version 4 a 9.3 for accuracy and an 8.8 for quality. Each grade was a judgment against her prompt, not a chance of being correct. A 9.5 did not mean the draft was 95% right. 3. **Catches.** Her notes listed 11 problems that existed before ChatGPT's review. Claude's self-review caught 5 of them and partly caught 2 more. ChatGPT caught 4 of the 7 that were still open when it reviewed. **The two reviewers saw different versions and different information, so this was not a controlled test.** 4. **Challenges.** Claude challenged a note that Nisha had added, because it could not confirm it, and she removed it. Claude also left out two true facts because it could not find them on the pages it checked. 5. **The last two grades.** Claude graded version 4 at 9.3 for accuracy and 9.0 for ease. After her edits, it graded version 5 at 9.2 for accuracy and 9.3 for ease. 6. **Limits.** When she asked, Claude listed the facts it could not fully check. Nisha confirmed 11 points against official sources and accepted the plain-English wording after her own review. 7. **The finish.** The draft never reached 9.5 in all three areas. She called it final anyway. > **Stop and think: a score is not proof.** A score is a reviewer's judgment. Evidence is what makes the judgment defensible. When reviewers disagree, Nisha must identify their assumptions, inspect their reasoning and verify the disputed claims. Claude's 9.5 and ChatGPT's 7 were both judgments. Neither one told Nisha which reviewer was right. Her fact check in step 6 did. ## A company of one Late that night, Nisha closed her laptop. She had a title, a short preface, a 25-chapter plan and 15 official sources. It had taken six versions, nine prompts and two subscriptions. It was a plan for a book, not the book itself. She noticed something else. She had not written most of the words. She had *managed*. She had given the brief, made the decisions, said no to extra work, asked for a second opinion, checked the facts and decided when the work was finished. To Nisha, that felt more like a manager's job than a writer's. People had told her that AI would replace accountants. Tonight it had not replaced her. Had it changed her job instead? If two AI Workers could help her plan a book in one evening, what else could they do? Review accounts for clients abroad? Convert company accounts from IFRS to US GAAP? Nisha began to dream of a company with one human and many AI Workers, with herself as its CEO. For now, it was only an idea. She remembered a line from her mentor: *"AI is not about building apps. AI is about solving problems, for people and for businesses."*[^practical-log] But she could not wait long. Her bills were still higher than her salary, and the two subscriptions now added about Rs 60,000 a month. By Monday morning she had to choose how to spend her evenings for the next three months. She could write all 25 chapters with the same six steps. She could change the method first. Or she could put the book aside and offer to convert a foreign client's accounts from IFRS to US GAAP, the first job for her company of one, which might bring in money sooner. Each path needed the same few hours each night. She could not do all three. Next, the [questions for discussion](questions-for-discussion.md). [^practical-log]: The counts in this list come from Nisha's AI Team: Practical Log, sections 3 and 4 and Appendices A to C.