The office gets an AI assistant

Chatbots and assistants

An AI chatbot answers questions in a conversation, by text or voice, like the help window on a store website that answers shipping questions at 2 a.m. An AI assistant goes a step further and does tasks for you: it sets reminders, sends texts, starts calls and adds meetings to your calendar, often by voice on a phone or computer.

Mr. Barkley: So the phone that sets my alarms is an assistant?

Theo: Yes. And the help window on our website is a chatbot.

Exam tip Chatbot = talks with you. Assistant = does things for you.

Generative AI

Generative AI makes new content from a written request called a prompt. AI-generated content includes text, images, audio and video, like a product description, an ad image or a draft email. AI-generated code is code an AI tool writes from a description, like 'write a function that adds sales tax to a price.' It saves time, but the result can be wrong, unsafe, or close to someone else's code that has license rules attached.

Theo: It wrote my coupon function in ten seconds. Then I spent an hour testing it.

Predictions and suggestions

AI predictions and suggestions learn from past data to guess what comes next: 'customers who bought this also bought,' a forecast of how much kibble each store will need next month, or a nudge that a customer's usual bag is probably running low. They don't create new content. They estimate and recommend.

Rosa: Could one tell me how much puppy food to stock for December?

Theo: That's exactly what prediction tools do. You still decide, though.

Use AI wisely

AI can sound sure and still be wrong. When it makes up a fact, that's called a hallucination. So check important answers against a real source. Have a person review and test AI-generated code before it goes live. Make sure AI-made images and text don't copy someone else's work. Never paste private data, like customer names and addresses, into a public AI tool, and use only the tools company policy allows.

Exam tip Treat AI like a very fast new intern: helpful and confident, and its work still gets checked.

Chatbots vs. assistants

An AI chatbot holds a conversation and answers questions, like the store's chat window that explains shipping and returns at any hour. An AI assistant works for one person and takes actions across that person's apps and devices: 'Remind me at 3 and text Maya I'm running late.' Both understand plain language. The difference is what they do: answer questions, or act for you.

Exam tip Answers in a chat window = chatbot. Sets alarms, sends texts, starts calls for you = assistant.

Generative AI: content and code

Generative AI creates new content from a written prompt. AI-generated content includes text, like a product description, and images, like an ad picture of a dog. AI-generated code is code written from a description, like 'a function that adds 8% tax to a price.' Before using either one:

• Check company policy and whether the content copies someone else's work.
• Review AI code line by line for bugs, security and licensing (it can look like open-source code with license rules).
• Test it with realistic data. The AI's own explanation is not a test.

Predictions and suggestions

AI predictions and suggestions use patterns in past data to forecast or recommend. 'Customers who bought this also bought...', a forecast of next month's kibble demand, a 'Time to reorder?' reminder, and a spam filter are all predictions. If a feature works out an estimate from history, it's prediction, even if it shows a message. It's generative only if it creates new content.

When AI makes things up

AI can state false information as if it were fact. This is called a hallucination. It can even invent sources, like a handbook section that doesn't exist. Always check important answers against the real source. A store chatbot should hand a chat to a human when it reaches a request it wasn't built for, like a refund, instead of inventing a policy.

Exam tip A very specific citation doesn't make an AI answer true. Check the real document.