Read a sample: 1. Getting Oriented: What Workplace AI Can and Cannot Do
A workday can fill up with small tasks that each seem manageable: read a long update, turn meeting notes into a list of next steps, write a courteous reply, and figure out why a spreadsheet formula is not doing what you expected. None necessarily requires a specialist. Together, though, they can consume the time you hoped to spend on the work behind them.
AI tools can help with some of that friction. You might ask one to suggest a structure for an update, turn rough notes into headings, or explain a spreadsheet formula in plain language. The result may give you a useful starting point in seconds. It may also leave out an important condition, assign a task to the wrong person, or produce a formula that works only if your sheet is arranged differently.
This book is about using AI as a practical assistant, not treating it as an authority. You do not need to know how to program or understand every technical detail behind the tools. You do need a clear task, appropriate information to work with, and a way to judge the result.
Before entering work information into any AI tool, check your employer’s rules and use only tools and data permitted for that purpose. A chat box that is easy to open is not necessarily an approved place for customer records, internal documents, or meeting notes. For the exercises in this chapter, fictional examples are enough.
The tools discussed in this book use a form of artificial intelligence commonly called generative AI. In everyday use, that means software that can generate content in response to an instruction. Depending on the tool, it may produce text, tables, formulas, or other material.
You give the tool a request, often called a prompt. It responds with an output. You can then ask it to revise the output, explain something differently, or focus on a narrower part of the task. This back-and-forth can feel like a conversation, but the tool is not a coworker who attended your meeting or understands your organization from experience. It generates responses based on patterns learned during training and the information available to it in the interaction.
That distinction matters. If you ask, “What did our team decide about the launch?” a tool cannot reliably know the answer merely because it can produce a convincing response. It would need access to relevant, permitted information, such as notes you provide or a document available through an approved connection. Even then, it may misunderstand the material.
Think of the output as a draft to inspect. Sometimes the draft will save considerable effort. Sometimes it will reveal that your request needs more detail. Occasionally, starting over without AI will be faster.
AI does not always appear in the same place. You may encounter a general-purpose assistant in a chat window, or you may see AI features inside software you already use.
A general-purpose assistant usually accepts a broad range of requests. You might paste a fictional paragraph into a chat and ask for a shorter version, then ask the same tool to explain a formula. Its flexibility can be useful when a task crosses several kinds of work. What it can access or do, however, depends on the particular tool and its configuration.
A document or note-taking feature may appear beside a document, within a menu, or as a command in an editing toolbar. It might offer to summarize selected text, rewrite a passage, or organize notes. Because it is near your work, it can feel especially convenient; the surrounding application does not, by itself, tell you what information the feature uses or whether your organization permits that use.
A messaging feature may help draft or revise a reply. For example, it could turn a few permitted points into a polite request for a meeting. You still need to supply the actual purpose and make sure the finished message says what you intend.
A spreadsheet feature may suggest formulas, explain functions, or help describe patterns in data. You might ask how to total values that meet a condition, then adapt and test the suggestion in a small fictional sheet. A formula that looks familiar can still refer to the wrong cells.
Some tools also accept files or connect to workplace information. Others work only with what you type or paste. Features, access, and settings vary, so the name or appearance of a tool is not enough to establish what it can do. Chapter 2 will help you compare options for a specific task rather than choosing one simply because it is available.
Generative AI is often good at producing smooth, confident language. That makes it pleasant to use, but it can also make mistakes harder to notice. A polished paragraph can contain a missing qualification just as easily as a rough one.
Imagine these fictional meeting notes:
> The team discussed moving the customer guide update to May. Priya will check the design schedule. No new deadline was agreed.
An AI-generated summary might say:
> Priya will complete the customer guide update by May.
The sentence is concise and plausible. It is also wrong in several ways: the notes say Priya will check a schedule, not complete the update; May was discussed, not agreed as a deadline. The AI has turned possibilities into commitments.
This kind of invented or unsupported detail is sometimes called a hallucination. The term…