The Context Vacuum

A transaction coordinator I know pulled up her email. Three in the afternoon. She’d asked AI to draft a follow-up to a buyer who was dragging on an inspection response. AI came back with something polite and generic — the kind of email that sounds like every real estate follow-up that’s ever existed.

She stared at it for about fifteen seconds. Then she opened the customer file. Read the last email thread. Saw what the buyer had actually said. Saw what the previous complication was. Saw what timeline pressure they were under.

Then she literally rewrote the whole email in about ninety seconds.

I asked her why she rewrote it.

She said “The AI’s version doesn’t know any of that. If I send that, I’m sending a response from someone who doesn’t know what’s actually happening.”

She hit send. Forty minutes later, the buyer responded.

Yeah. Here’s the thing — that ops person wasn’t fixing a broken AI. She was fixing an incomplete handoff. And she’s doing it five times a day without naming it.

What the AI actually doesn’t know

Quick name-check first. We wrote a piece called Stop Hiring AI. Start Building It. that introduced the Professional Recipe — seven things any operator with sense gives a new hire on day one. One of those seven is Context — Ingredient #2.

Here’s what Context is, and why it matters.

Training (Ingredient #1) is what’s true about your business every single day. You’re a real estate office. Your clients buy and sell homes. You close deals at a specific pace. You have a tone. You have standards.

Context is what’s true right now, about this situation. This buyer has been dragging on the inspection report. Appraisal came in low. They’ve got financing contingencies we’re watching. You know all of this because you’ve read the thread. AI hasn’t.

Here’s the gap that kills the draft. Most operators give AI training (the persistent setup, the system prompt) and then just ask the question. No context. AI is running on what it knows about your business in general, but nothing specific to this moment.

So it writes a follow-up that’s correct, grammatically clean, and completely divorced from what’s actually happening.

You can feel the wrongness immediately. Your ops person can. Every agent in your office can. So they rewrite it.

Right? And that’s the thing — they rewrite it in private. Every rewrite is evidence that the AI needed context it didn’t have. Every rewrite is a lesson that never gets fed back to the system. AI keeps drafting context-less followups. Ops person keeps silently fixing them. Nothing improves.

The habit you already have — but with people

Here’s what snaps this into focus. You would never send a junior agent to draft a client email without the file.

That junior agent sits down. You hand them the customer history. The previous emails. The current status. The specific negotiation point you’re stuck on. The pressure — time pressure, financing pressure, whatever it is. You say “Based on all of that, draft a follow-up that moves this forward.”

Junior agent reads the file. Understands the stakes. Drafts something that lands differently than if they’d just been told follow up with this buyer.

That’s context. You’re not leaving it out because you don’t know better. You’re not leaving it out because you’re cheap. You do it because you know the output depends on it.

So why are you leaving it out with AI?

What’s different about AI

Here’s where it gets interesting. That junior agent, after you hand them the file once, internalizes the situation. They develop judgment about what matters. A week later, they’ll remember oh, this buyer was sensitive about timeline pressure without you re-explaining it.

AI doesn’t do that. AI has no memory of previous files you’ve handed it. Every task starts at zero. You hand it the context. It reads it. It produces output. Then the context is gone. Next task, you hand it the context again.

This isn’t a limitation. It’s actually the design. But it’s a design that means context can never be optional. You skip it with a person, they learn. You skip it with AI, it reliably produces the wrong output. Does that make sense?

Most operators don’t see this as I need to change how I work. They see it as AI isn’t good enough. Then they walk away from the tool, frustrated.

How to feed the AI the file

Here’s the move, and it’s simpler than you think.

For any task where the output quality matters — a follow-up email, a property description, a communication about a deal complication — build the habit of pasting in the context first.

What is context in real estate? It depends on the task.

For a buyer follow-up. Buyer’s name. What they’re under contract for. What stage the deal is in. What the last conversation was. What’s holding it up. What timeline pressure exists. What you know about their sensitivity (they hate being rushed, or they’ve got a hard deadline, or they need frequent updates).

For a listing description. Property type. Neighborhood. Specific features. List price strategy. Seller’s timeline. Whether there are quirks the listing needs to address (small lot, older systems, special features).

For a negotiation message. What’s been offered. What’s been countered. Where the gap is. What leverage exists. What the other side is likely thinking.

You paste all of that in. Then you ask the task. “Based on the above, draft a follow-up that [accomplishes X].”

Output changes. Dramatically.

That ops person’s rewrite wasn’t magic. She just had the file. She knew what was actually happening. That’s what made the email land.

Why this matters for the ops person

Here’s the thing — you’re the person in the office who knows the context. You’re reading the threads. You’re coordinating the moving parts. You’re the person who catches when something doesn’t add up.

That knowledge is the most valuable input the AI has access to.

When you skip feeding context, you’re not just getting a mediocre draft. You’re leaving the AI operating on surface-level information while you’re sitting on situation-specific intelligence that would change everything. You have the file. You’re just not sharing it.

Ops person becomes the one silently fixing AI outputs instead of the one leveraging AI to multiply their own work.

The Monday Move

So. Pick one customer-facing task this week. Doesn’t have to be big. A follow-up email. A status update. A showing coordination message. Something where the current version isn’t quite landing.

Before you ask AI to draft it, paste in the context. History. Last conversation. Current status. What matters right now.

Then ask the task.

Compare that output to what you would have gotten cold. See what changes.

My bet — you’ll see the difference by the second or third task. Once you see it, you can’t unsee it. And once you build the habit, the AI stops sounding generic and starts sounding like someone who actually knows what’s happening.

The shift

AI is a confidently wrong stranger when you don’t feed it the case file. You already know this — that’s why you’re rewriting everything. What you might not have named is that the wrongness isn’t because the AI is broken. It’s because you’re asking it to work without the information that would make it right.

You wouldn’t do it to a person.

So. Feed the file.

That’s the whole shift.

Framework: The Professional Recipe — Context ingredient (#2). Related failure modes: The Context Vacuum — context discipline broken at the workflow level.

Companion piece: Stop Hiring AI. Start Building It. — the parent framework. This one closes the gap on Ingredient #2 specifically.