Chapter 1. From Chatbots to AI Workers

The point

AI text boxes look the same today as they did in 2023, but what happens behind them has changed completely. In 2023 the box gave you back text — an answer you copied somewhere else yourself. Today it can give you back finished work: real documents, spreadsheets, and presentations that it can also open again and edit. After this chapter you can explain that change. You can look at any AI product and place it on a simple four-rung ladder — from tools that answer a question to tools that finish a job. You can keep three ideas separate that product marketing blurs together: whether the AI remembers and continues its work (continuity), which computer the work happens on (where it runs), and where the finished output ends up (where it is kept). Maria's story below shows what each of these costs when you get it wrong. And when the next big AI announcement lands, you can tell what is genuinely new from what is a new name for the same thing.

Why it matters

Maria runs the office at a wholesale supplier in Ohio with about 40 staff. On Monday she opens her AI assistant and types one line: "Summarize these vendor invoices." She attaches 15 PDFs. A tidy paragraph comes back. She copies the totals into a spreadsheet by hand, checks two of them against the PDFs, and spends another 40 minutes copying the rest. She uses the tool exactly as she learned to in 2023.

The same box could have done the whole job. If she had asked for the finished result — the spreadsheet itself, not a summary — it could have read every invoice and built it. It could have checked each total and flagged the one invoice that was billed twice. That duplicate was for $4,850. Maria paid it.

On Wednesday her manager tells her to "let the AI do it." This time she asks for a report of unpaid vendor bills by how overdue they are. The assistant builds a spreadsheet and shows her a short summary in the reply. She reads the summary and closes the tab. The next Monday she asks for "last week's spreadsheet, updated." The summary is still in the conversation. The spreadsheet is not. It was built in a temporary workspace and was never saved in a place where she could open it again.

Maria made two common, expensive mistakes, and neither was about intelligence. First, she got 2023 value from a 2026 system, because nothing on the screen told her the box had changed. Second, she assumed that because the conversation remembered her, the work survived too. The continuity was real — the summary was right there. But she did not know where the work had run, or where its results were kept, and those are different questions. This chapter teaches you to ask all three.