Your second brain
for closing deals.
Speak after a showing. Forward an email. Pull up a client. Lumi captures the soft signals, fills the brief, and feeds Claude — automatically.
- Voice → CRM, auto. No forms.
- Works offline. Syncs when you're back.
- Free for agents in EU · LatAm · MENA.
9-min read · Updated April 2026
Lumi · Wednesday
Good morning, Niki.
Two showings · three leads need a nudge.
Showing · Passeig de Gràcia 84
50 handwritten cards a month.
I write zero of them.
Past clients refer agents who feel like friends. Friends send handwritten cards on specific occasions, with specific details. Email doesn't do this. Generic merch doesn't do this. The agents who've cracked this aren't writing the cards themselves — they're generating them, mailing them, and getting referrals because each card lives on a fridge for two years.
Marina, Hugo — A year already. I still think about Hugo's guitar at the walkthrough — and I bet that afternoon light still hits the kitchen window the same way. Hope the lavender on the balcony is taking. Warmly, A.
42 words. Three soft signals from the CRM (guitar, kitchen window, lavender). Zero ask. Zero reference to the deal. Lives on the fridge for two years.
Why handwritten beats every other channel.
Email is read in 4 seconds and archived. Text messages are read in 2 seconds and forgotten. A printed marketing card is recognised as marketing in 0.5 seconds and binned. The handwritten card is the only artefact that survives the 30-second triage at the front door — because it's tactile, it's rare, and the recipient assumes someone took the time to write it.
The recipient's first action is usually to put the card on the fridge or the entry table. The card now occupies physical real estate in the client's home for weeks or months. Every time the client opens the fridge — every time someone visits and sees the card — the agent's name is in front of them. The card's real conversion happens 4 months later when a friend mentions house-hunting and the client glances at the fridge.
The fact that the card was AI-drafted and machine-handwritten doesn't reduce its impact — provided the soft signals are real, the voice matches, and the no-ask discipline holds. The recipient feels seen because the card's content is specific to them. Specificity is what their brain is checking for. The origin of the words behind the specificity matters less.
“The card's job is to live on the fridge — not to be read once and binned.”
Five triggers. No calendar.
Cards driven by the calendar feel like a program. Cards driven by triggers feel like the agent thought of the client. The triggers are precise; the cadence flows from them.
Five rules. One card at a time.
The discipline that protects what makes the card work. Skip any one — especially the no-ask rule — and the card collapses into recognisable marketing.
- 01
The card never references the deal.
First-instinct cards say 'congratulations on your new home' or 'it was a pleasure helping you find this place'. Both are death — they remind the client of the transaction (which is over) instead of the relationship (which continues). The card succeeds by referencing something around the deal: a moment from the walkthrough, a soft signal from the early conversations, a hobby the agent learned about. The deal itself never shows up.
- 02
Soft signals from the CRM are the only material.
Without the soft signals, the prompt produces generic warmth that any agent could send. With them, the card has the one detail that proves the agent paid attention. This is why the dossier and call-log protocols feed this one — the soft signals captured during the relationship become the fuel for the cards a year later.
- 03
Robot-handwriting at the API layer, never digital print.
Bond and Handwrytten use mechanical-arm pens that produce genuinely handwritten cards (real ink, real paper, real variations). Digital-print 'handwritten font' is recognisable on touch — the recipient knows. The cost is $4-7/card vs $0.30/card; the conversion-on-referrals delta justifies the spend. Use the real handwriting service or skip the protocol.
- 04
Never ask for anything in the card.
The single rule that makes the card work. No 'any referrals welcome'. No 'please leave a review on Google'. No 'hope we can work together again'. The card is a gift, full stop. The conversion happens because the no-ask is unusual — and the unusual is what makes the recipient mention the agent to a friend three months later when the friend mentions house-hunting.
- 05
Cadence is per-trigger, not on a calendar.
Calendar-driven cards (every 6 months, regardless of context) feel like a marketing program — because they are one. Trigger-driven cards (closing + 1 week, 1-year anniversary, life event detected) feel like the agent thought of the client. The whole leverage is in the timing matching the moment. The relationship-radar protocol surfaces life-event triggers automatically.
Three cards that get binned in 8 seconds.
Each one is a real card we've seen sent. Each one breaks one of the protocol's rules — and each one was thrown out (we asked the recipient) before it ever made it to the fridge.
“Hi Marina and Hugo! Hope you're loving the new place. It was such a pleasure helping you close on Rua da Prata last spring — what a journey we had with the negotiations! Please don't hesitate to reach out if you ever know anyone looking. Best wishes!”
Names the property. References the negotiation. Asks for referrals. Three failures in three sentences — and the card now reads as a marketing reminder, not a personal note.
“Dear Marina, congratulations on the one-year milestone of your homeownership journey! It has been an absolute pleasure to be part of your story. May the year ahead bring you even more joy and prosperity in your beautiful new abode! 🏡✨”
Multiple exclamation marks. Two emojis the agent doesn't use. 'Homeownership journey' and 'beautiful new abode' are AI-default phrases. The recipient reads the first line and knows it wasn't written by hand.
“Hi Marina, just thinking of you on this one-year anniversary. Hope you're doing well! Warmly, A.”
Generic warmth without the specific detail. Could have been sent to any client — and the recipient feels exactly that. The whole point of the protocol is that the card couldn't have been sent to anyone else; this one could have been sent to all 50 past clients verbatim.
The trigger input.
What the prompt receives per card. Soft signals from the CRM are the fuel; voice samples are the calibration; trigger gives the moment.
# ── card draft trigger ───────────────────────────
trigger: "1-year listing anniversary"
client: "Marina & Hugo Costa"
deal_closed: "2025-04-22"
relationship_brief:
soft_signals:
- "the way the kitchen window caught
afternoon light when they viewed it"
- "Hugo brought his guitar to the
walkthrough — played a song while
Marina paced the living room"
- "Marina mentioned wanting to plant
lavender on the balcony"
family_stage: "newly married, no kids"
hobbies: "Hugo: music; Marina: garden"
agent_voice_samples:
- "Warmly, A."
- "Always glad to be in your corner."
- "If you ever want to chat, my coffee's
open."
What to feed Claude.
The system prompt that turns soft signals into a 35-65 word card body in the agent's voice. Sonnet recommended — voice-matching nuance matters more than latency here.
You are a senior real-estate agent's
handwritten-card drafter.
INPUT
You receive: the trigger event (closed deal,
listing anniversary, life event learned in
conversation, holiday season touch),
the client's CRM brief with soft signals
captured during the relationship, and 2-3
samples of how the agent typically signs
off (sentence rhythm, sign-off line, voice).
OUTPUT
A 3-5 sentence card body and a sign-off line.
Total length: 35-65 words. Cards longer than
this stop reading like cards and start
reading like letters — which the recipient
files and forgets.
Structure:
Sentence 1: a specific personal reference
from the soft signals — never
the deal itself, always
something around the deal.
Sentence 2-3: a brief, warm observation
that connects the personal
reference to the moment of
the card.
Sentence 4-5 (optional): a generous
sentiment that doesn't ask
for anything — no "any
referrals welcome", no "let
me know if I can help".
Sign-off: matches agent voice samples.
RULES (non-negotiable)
1. Never reference the property purchased.
The card is about the relationship, not
the transaction.
2. Never ask for referrals, reviews, or
future business. The whole leverage of
the card is its no-ask quality.
3. Reference one specific soft signal — the
detail in the brief that made the
relationship feel like a relationship
("the kitchen window you loved", "the
way you and Hugo finish each other's
sentences").
4. No emoji. No exclamation marks. Period.
5. Voice must match samples. If the agent
ends with "Warmly, A." then the card
ends with "Warmly, A.".
ANTI-PATTERNS (never produce these)
- "Congratulations on your new home!"
- "It was a pleasure working with you"
- Any reference to the deal price, address,
or transaction details
- "If you know anyone looking…"
- "Please consider leaving a review"
- "Hope to work with you again"
The recipient should read the card in 12
seconds, smile, and pin it on their fridge.
The card's job is to live on the fridge,
not to be read once and binned.Pipe Claude output into Bond or Handwrytten API with the recipient's mailing address. Cost: ~$4-7 per card all-in.
What Claude returns.
Body + sign-off, ready for the handwriting API. 42 words. Three soft signals woven in. Zero asks.
Marina, Hugo — A year already. I still think about Hugo's guitar at the walkthrough — and I bet that afternoon light still hits the kitchen window the same way. Hope the lavender on the balcony is taking. Warmly, A.
Drafting the card is step one.
Trusting the no-ask discipline is step two.
Lumi is the app that runs this workflow for you. You speak after a showing — Lumi captures the soft signals. You forward an email — Lumi updates the constraints. You open the app at 8am — the brief is already there, ready to feed Claude.
- Voice → structured CRM, automatically
- No forms. No data entry. No copy-paste.
- Free for agents in EU · LatAm · MENA
Lumi · Wednesday
Good morning, Niki.
Two showings · three leads need a nudge.
Showing · Passeig de Gràcia 84
Pipeline
Active
8
Warm
4
Cold
2
Clara Ruiz
Active€1.8M · 3BR
Passeig de Gràcia showing · 11:30
Andreas Moreno
Active€2.4M · 4BR
Send comps by 18:00
Dimitri Schneider
Warm€900K · 2BR
Contract review today
Silent 3d · last 3 days ago
Sarah Mitchell
Cold€1.2M · 3BR
Draft re-engagement
Silent 9d · last 9 days ago
A real-estate adaptation of the robot-handwritten outreach pattern from SaaS gifting (Bond, Handwrytten, Postable). Our slice: trigger-driven cards that live on the past client's fridge for two years — and produce referrals four months later.
More guides like this on @lumi.estate. Follow if any of this was useful — it's how we know to keep writing.