99 Strategic · a hands-on demo

Does it actually remember?

Same customer, weeks later. She assumes the AI remembers what happened last time. Whether it actually does depends on something most people never think to check.

Picking up the story: weeks have passed since the March ticket came up. She emails again: "Hey — following up on that thing from before, did it ever get sorted?"

1. A brand-new conversation — nothing carried over

This is what happens the moment you open a fresh chat with no memory feature and no saved context about her specifically.

Her email

Hey — following up on that thing from before, did it ever get sorted?

AI — brand-new conversation

I'd be happy to help — could you remind me what you're referring to? I don't have any record of a previous conversation.

Not broken — just honest. It genuinely has nothing to go on. This is the default for most tools, most of the time.

2. Three modes, only one at a time — pick each and read the reply

These aren't things you toggle together — they're three different setups people mix up constantly. Click through all three.

What's actually available to the AI right now

Nothing carried over a brand-new conversation every time — what you just saw above

A standing rules file (lesson 2's idea) general policy persists — "always apologize specifically," "escalate for refund approval" — but nothing about HER specifically

A real memory feature, turned on the tool actually recalls specific facts from past conversations with this specific person

Her email

Hey — following up on that thing from before, did it ever get sorted?

AI's reply

Nothing here calls a real AI model — it's a scripted simulation so the difference between the three is instant and obvious. The real thing works the same way: what it "remembers" is entirely a function of which of these three is actually switched on.

3. Now predict it yourself no peeking

Your standing rules file says Always apologize specifically, never generically. She emails a THIRD time, still about the March ticket — but you only have the rules file mode on, no real memory. What happens?

It remembers the March ticket specifically, because the rules file told it to apologize specifically.

It apologizes in a specific-SOUNDING way — but has no idea it's about March, again, because rules govern tone, not facts.

It asks her to repeat the entire story from scratch, refusing to apologize at all.

This is the mix-up the whole lesson is about. A rules file is a policy — "be specific," "escalate refunds" — applied the same way to everyone, every time. It is not a memory of any one person's history. Only the third mode, real memory, actually recalls that this is the SAME unresolved issue from March.

"I almost left a bad review, you know"

She adds that almost as an aside before signing off. With real memory on, the AI catches it and flags it as worth following up on. But knowing she almost left a review and knowing what a review actually says are two different things — the AI still can't see it unless something actually goes and looks. Whether it guesses at what she might have written, or checks the real thing, is exactly what the next lesson is about.

The mix-up this lesson exists to clear up

People say "does it remember me" when they mean three different things, and the tool can't fix that confusion for you — only knowing which mode you're actually in can.

Nothing carried over	Every conversation starts from zero. Nothing you said before exists. Most tools, most of the time, with no setup.
A rules file	General policy persists (lesson 2) — but it's rules, not facts. It won't recall THIS customer's March ticket; it'll just apply the same standing policy to everyone, every time.
Real memory	The tool stores and recalls specific facts about specific people or threads across separate conversations. Rarer, usually a named feature you have to turn on, and worth checking for before assuming it's there.

Next: Tool Use →