Lesson 1 of 8 · 8 min read · last verified 2026-08-26
It arrived without asking
In this lesson you will:
- Recognise where AI has been added to tools you already use
- Explain why unchosen AI needs different scrutiny from chosen AI
This whole curriculum has assumed you opened an assistant on purpose. For most people that is not how it started.
It started with a suggested reply appearing under an email. A summary at the top of a search page. A phone keyboard finishing sentences. A “write with AI” button in a document. Nobody chose any of it — it turned up in software they already had, usually switched on.
Chosen and unchosen are different situations
When you open an assistant, you know it is one. You bring expectations. You know to check things, because eleven modules told you to.
Embedded AI removes all three conditions:
You may not know it is AI. A “suggested” reply, a “smart” summary, a search result that reads like an answer. The word AI often appears nowhere.
You never evaluated it. E2 taught a fair comparison before choosing. Nobody ran that on the summariser their email provider enabled last Tuesday.
It acts by default. Not when you ask. Continuously, on everything, unless you find the setting.
That combination is why this module exists. Everything you have learned still applies — it just has to be applied to something you did not go looking for.
Where it already is
Not exhaustive, but enough to recognise the pattern:
Email — suggested replies, tone rewriting, thread summaries, priority sorting.
Documents and spreadsheets — drafting, rewriting, formula generation, summarising.
Search — generated answers above the results. For many people this is now their main AI use and they have never called it that.
Phones — keyboard prediction, photo editing and search, voice assistants, call transcription, on-device summaries.
Messaging — suggested replies, translation, chat summaries.
Customer service — the first responder is frequently a model.
Work software — meeting notes, ticket routing, CV screening, performance tooling.
The default is the decision
A quiet but important point.
Most people never change a default setting. So whoever picks the default is effectively deciding the behaviour for almost everyone — and defaults are chosen by the vendor, in their interest, which is usually “on”.
That is not sinister; it is how software has always worked. But it means “I did not turn this on” and “I agreed to this” are different statements, and only the second one implies you considered it.
L3 is about finding those switches. L2 is about noticing there is something to switch.
What does not change
Reassuringly, the substance transfers directly:
- E7 still applies. A generated search summary needs the same checking as a chatbot answer, and gets less of it because it looks like a search result.
- E4·L8’s paste test still applies. An email assistant reads your email. A meeting summariser hears your meeting.
- E9’s caution still applies. If a captioning feature is doing the work, know its failure modes.
You are not learning a new subject. You are pointing an existing one at something that did not announce itself.
Try it now (5 minutes)
Open the app you use most for email or messaging. Look for anything offering to write, summarise, suggest or organise.
Count them. Most people find between two and five, and are surprised by at least one — that surprise is the point of this module.
Check your understanding
Recap
Most people’s AI arrived as a feature in software they already had, switched on by default. That removes the three things that make deliberate use safer — knowing it is AI, having evaluated it, and it acting only when asked. Everything you have learned still applies; it now has to be applied to something that did not announce itself.
🗂 3 flashcards from this lesson join your daily review.
Next: Spotting it