Prep for every meeting in 10 seconds

It is 1:58pm on a Tuesday. You are walking to the conference room. Your 2pm with Sarah Chen at Acme Corp starts in 90 seconds. You had back-to-back calls all morning. You have not looked at the Acme deal since last week. You know the broad strokes, but the details are fuzzy. What did she say on the last call? Did you send that document you promised? Who else has been looped in on their side?

You pull out your phone. You open Chief of Staff.

"Prep me for my meeting with Sarah."

What Comes Back

In under a second, Chief of Staff responds. Not with a generic summary. With the specific context you need for this meeting, right now.

The deal: Acme Corp, $240K, Stage 3 Negotiation. Close date June 30.

Your champion: Sarah Chen, VP Engineering. She brought you in after seeing your keynote at SaaStr. She has been your internal advocate since the first call.

Last call, Tuesday: Sarah raised procurement timeline concerns. Her CTO Jake Martinez was on the call but barely spoke, which is unusual because he was talkative in the first two meetings. You committed to sending the implementation guarantee document by Thursday. You did not send it.

Recent email activity: Jake forwarded your pricing proposal to their CFO on Wednesday. No response from the CFO yet. That forwarding happened without you being CC'd. Sarah mentioned it in a one-line email on Friday.

Calendar context: Sarah has a board meeting next week. If you can get CFO alignment before that meeting, the deal has a real shot at closing in June. If the board meeting passes without your deal on the agenda, you are looking at a July or August close at best.

Suggested play: Lead with the implementation guarantee you owe her. Acknowledge that it is late. Ask about Jake directly. He went quiet on the last call and his forwarding the proposal to the CFO without looping you in could mean he is either championing it internally or building a case against it. Float a mutual close plan with specific dates tied to the board meeting.

Where That All Came From

None of that required a server. None of it required an internet connection. Here is what happened in that one second between your question and the answer.

Your question was embedded into a 384-dimensional vector by the on-device embedding model. That vector was compared against the 50,000+ chunks in your local sqlite-vec database. The search returned the most relevant chunks in about 1 millisecond: the HubSpot deal record for Acme Corp, three sections from Tuesday's Zoom call transcript, the email thread where Jake forwarded the proposal, Sarah's calendar entry for next week's board meeting, and your task list showing the implementation guarantee doc you never sent.

Those chunks were assembled into a prompt and fed to the on-device LLM. The model generated the response. Total time from question to answer: under one second. Total data sent to the cloud: zero bytes.

What Life Looks Like Without This

You know what the alternative looks like because you live it every day. You open HubSpot on your phone. The mobile app is slow. It takes 8-10 seconds to load. You search for Acme. You tap into the deal. You scroll through the activity feed. Half the notes are incomplete because your reps log the minimum. The last note says "good call, follow up next week." That tells you nothing.

You try to remember the last call. You know Sarah said something about procurement. Was it timeline? Was it budget? Was it a specific objection? You cannot remember. You open Zoom to find the recording. The search is bad. You cannot find the right call in 90 seconds.

You check email. You have 47 unread messages. You search for "Acme." Six threads come up. You tap through them looking for the latest. You find Sarah's email about Jake forwarding the proposal. But you do not have time to read the full thread.

You walk into the meeting half-prepared. You fake confidence. You ask questions you should already know the answers to. Sarah notices. She does not say anything, but she notices. And the deal moves a little slower because of it.

Other Moments This Changes

Meeting prep is the obvious use case. But once your CRM data lives on your phone with an AI that can actually read it, other moments start to change too.

You are in the car on the way to a client dinner. "What is my pipeline looking like this month?" You hear the answer through your car speakers. $1.2M in Stage 3+. Three deals at risk. One deal you have not touched in 11 days. The AI names the deals and the risk factors. You did not open an app. You did not take your eyes off the road.

Sunday night, before the week starts. "What does my week look like?" Chief of Staff walks through your calendar, meeting by meeting. Monday is internal. Tuesday you have the Acme follow-up and a first call with a new prospect. Wednesday is demo day, three back-to-back demos. Thursday is forecasting with your manager. For each meeting, it gives you the 30-second version of what you need to know. Your whole week, contextualized, in about two minutes of voice conversation while you are sitting on the couch.

After a call, walking back to your desk. "Log a note on Acme: Sarah confirmed budget approved, procurement starts next week, next step is security review with their CISO." That note is saved to the local database with the timestamp, linked to the Acme deal, and available the next time you or the AI references this account. You did not open HubSpot. You did not type anything. The CRM is updated before you sit down.

On a flight, no Wi-Fi. "Draft a follow-up email to Sarah referencing our conversation today." Chief of Staff writes it using context from the deal record and the note you just logged. You review it, edit one line, and queue it to send when you land. All offline. All on-device.

The Compound Effect

One meeting where you walk in prepared does not change your number. But every meeting, every day, for an entire quarter? That changes outcomes. The rep who walks into every conversation knowing the deal stage, the last call context, the open commitments, and the right next play is the rep who closes. Not because they are smarter. Because they never walk in cold.

I have carried a quota. I know what it feels like to walk into a meeting and realize you forgot a commitment you made two weeks ago. I know the sinking feeling when a champion asks about something you should have followed up on and you have to tap-dance. Those moments erode trust. They slow deals down. And they are entirely avoidable if you just have the right context at the right time.

This is not a demo scenario. This is what happens when your CRM data lives on your phone and you have an AI that can actually read it. Ten seconds of prep. Every meeting. Every day.