Tagline (gold-shimmer treatment, DM Serif Display 56px, sits above H1):
Learn, build, accelerate.
Your CRM doesn't rank your sphere. ChatGPT does — for $20.
The CRM sorts alphabetically. The dashboard sorts by last login. Neither one tells you who's actually warm. ChatGPT does — in 90 seconds, on a CSV you already own, for $20 a month instead of $700.
This page is the prompt. Copy it, paste it, run it Tuesday morning before the kids are out the door.
What this prompt does
You paste it into ChatGPT. You attach your sphere CSV. You get back a ranked top 20 with a 30-second opener and a why-now reason for each contact.
Outcome. A call list ranked by deal-readiness, not by alphabet or last login. The opener is written for each contact based on what's in their notes — the kid graduating in May, the divorce mentioned in October, the moving-company referral last spring. The why-now reason is the line you say to yourself before you dial, so the call has a point.
Cost. $20 a month for ChatGPT Plus, Claude Pro, or Gemini Advanced. Pick one.
Time. Four minutes including the CSV export. Three minutes after the first run, once the workflow's automatic.
Replaces. The $700-a-month predictive vendor your CRM keeps trying to upsell. See the SmartZip vs foundation model breakdown for the cost-and-outcome math.
The prompt
Copy it verbatim. Don't soften it. The constraints do the work.
You are an operator's call-prep assistant. I'm a working REALTOR.
I'll attach my sphere CSV — name, email, phone, last contact date,
notes field, transaction history.
Rank the top 20 contacts I should call this week, scored on:
1. Last contact date — anyone over 6 months gets a recency boost.
2. Life-stage triggers in the notes field. The triggers that count:
- kids (graduation, college, new baby)
- divorce or separation
- retirement or layoff or job change
- moving-company referral or relocation mention
- inheritance or estate work
- downsize or upsize signals
3. Prior transaction with me — past clients beat strangers.
4. Referral history — anyone who sent me a deal in the last 24
months ranks higher.
Skip anyone I closed in the last 18 months UNLESS they referred
someone since.
For each of the 20, output:
- Rank (1-20)
- Name
- Why-now reason in one sentence — the specific signal you found
- 30-second opener I can read off the screen, written in plain
spoken English, contractions, no exclamation marks, no
buzzwords. Sound like an agent who knows the person, not a
call-center script.
After the list, flag separately:
- Any contact I haven't called in 6+ months whose notes contain
a life-stage trigger but didn't make the top 20.
- Any contact whose notes contradict each other (e.g., "moving
to Brentwood" + "loves Cool Springs, won't leave"). I'll
decide which is current.
Do not invent triggers that aren't in the notes. If a contact's
notes are sparse, rank them lower and say so. Do not add emojis.
Do not editorialize.
That's the artifact. The rest of this page tells you how to run it.
How to use it (5 steps)
Step 1 — Paste the prompt into ChatGPT
Open a fresh chat. Don't load last week's CMA prompt into context. Paste the prompt above verbatim.
If you're on Claude Pro, the same prompt runs cleanly. If you're on Gemini Advanced, the same prompt runs cleanly. The artifact is model-agnostic.
The privacy toggle on the paid tiers — off by default for training. Confirm it in settings the first time. Move on.
Step 2 — Attach your contact list CSV
Open your CRM. Follow Up Boss, kvCORE, Top Producer, Wise Agent — all of them export. File menu, Export, CSV. Include name, email, phone, last contact date, notes field, transaction history.
The notes field is the one that does the work. It's where the messy half-thoughts live — "kid's graduating in May," "mentioned divorce on the phone in October," "moved her aunt to Westmoreland." If your CRM gives you a choice, export everything. The model can ignore noise. It can't read what isn't there.
Drop the CSV into ChatGPT — paperclip icon, or drag the file onto the chat. Upload is cleaner for files over a few hundred rows. Paste works fine for smaller exports.
Step 3 — Read the output
ChatGPT runs the prompt in 90 seconds. You get a numbered list — top 20 contacts, ranked. Each row has a why-now reason and a 30-second opener.
Read it once start to finish before you edit. Resist the urge to argue with rank #3 in the first 10 seconds. The full list tells you whether the ranking logic held — if 17 of 20 names look right, the model read the data. If 12 of 20 look wrong, your notes field is sparser than you thought, and that's the bigger problem.
For the how-to walkthrough on a real CSV — the same one we run live in the Cool Springs workshop — see the full version with screenshots.
Step 4 — Validate the top 5
Check the top 5 against your memory. The model is reading notes — you're reading the human.
If a name surfaces and you remember why-now is wrong — the divorce settled six months ago, the kid graduated, the move already happened — type it back into the chat. "Update for #3: the divorce closed in March, they're stable now. Re-rank." The model adjusts the rest of the list.
This is the OODA loop in production. Observe the output, orient against your knowledge, decide whether to ship or correct, act. Ninety seconds total. Catches the hallucinated reason before it lands in a call.
Step 5 — Call the top 5 this week
Block five 12-minute calls across the week. Tuesday morning is the slot most agents miss — that's why the workflow runs Tuesday before showings start.
First call uses the opener verbatim. Read it off the screen. After two or three calls you'll edit the opener to match your voice — that's expected. The opener isn't a script. It's a starting line.
Track which calls turn into appointments. After four Tuesdays you'll know which life-stage triggers in your specific sphere actually convert. Tighten the prompt around those. The system gets sharper every week.
Why this works
Two reasons. Both come from primary sources.
Reason one — foundation models read weak signals a CRM dashboard can't. Andrej Karpathy framed this on X in 2025 as Software 3.0. English is the new programming language. The model is the runtime. You don't structure the data first and query it after — you hand it the mess and ask the question.
A CRM dashboard only knows what got typed into a status field. If "kid's graduating in May" sits in the notes column instead of a structured "life event = graduation" dropdown, the dashboard can't see it. The dashboard's blind by schema. The foundation model reading the prose isn't.
Reason two — prioritization beats volume on owned data. The Lead Response Management Study (Oldroyd, MIT/InsideSales — 15,000 leads, 100,000 call attempts) measured what happens when a rep replies in 5 minutes versus 30. Five-minute response wins 21x more qualified deals. Industry average response time is 42 hours.
The same logic governs sphere calls. Getting to the right person at the right window matters more than getting to a hundred wrong people. The bottleneck isn't volume. It's knowing who to call, in what order. The Tuesday-morning workflow and the 5-minute response workflow are two halves of the same operator-grade pipeline.
The deeper layer — Eugene Yan's roundup of working LLM patterns (O'Reilly, 2024). Foundation models earn their keep on small tasks with structured input, structured output, bounded scope. Ranking 487 contacts against five named criteria with a defined output format is exactly that shape. It's not RAG, it's not fine-tuning, it's not an agent loop. It's a prompt on a CSV. The simplest pattern that works.
NAR's primary research backs the volume math. The 2025 Member Profile — 66% of sellers find their agent via referral or prior transaction. 20% of a typical Realtor's income comes from past clients. The signal is in the sphere. Ranking it is what's been missing.
When to upgrade
The Tuesday-morning prompt isn't a universal answer. There's a real workflow where paid predictive infrastructure earns its slot. The line is mechanical, not opinion.
Three gates. Pass all three and a paid tool starts paying for itself.
Gate one — volume. 30+ closed deals per year, or $300K+ gross commission income. Below that, a $6K-a-year subscription eats 10–15% of gross before a single deal closes.
Gate two — labor. A salaried ISA, a second licensed agent, or a transaction coordinator answering inside the 5-minute LRM window. Predictive output is worthless without a body to action it. Sole proprietor with showings on the calendar fails this gate by definition.
Gate three — geographic concentration. 8+ transactions per year in a single ZIP or sub-market. Predictive farming only ROIs when you can door-knock the same households for 12+ months. Bounce ZIPs and the data turns over before you do.
Pass all three and the foundation model still helps you think — but you also need predictive infrastructure under it, because manually scoring a 5,000-household farm against life-stage triggers stops being a 4-minute Tuesday job. At that volume, look at Structurely or the comparable enterprise stack, and pair it with the Context Card system from the Listing Machine.
Pass two gates and it's marginal. Pass one or zero and the prompt beats the vendor on every axis — cost, contract length, signal quality, exclusivity.
For the median 12-deal sphere-driven REALTOR — the agent at Compass Hendersonville with 487 contacts in Follow Up Boss — the prompt wins. Run it weekly. Bookmark this page.
What this replaces
The vendor stack the prompt cuts down to $20 a month.
| Tool |
Price/mo |
What it claims |
What the prompt does |
| SmartZip |
$700 |
Predictive seller scoring on a list of strangers in your ZIP |
Ranks the warm contacts already in your CRM |
| Offrs (exclusive) |
$600 |
Exclusive seller scores in your ZIP |
Same — on owned data with notes the vendor can't see |
| Catalyze AI |
$360 |
~30 inherited-property leads/mo, 50-mile radius |
Surfaces inherited-property triggers from your own notes field |
| Likely.ai |
not published |
Pricing gated behind sales |
A pricing-gate is a tell |
| ChatGPT Plus |
$20 |
A foundation model |
The actual workflow on owned data |
The vendor list buys you a list of strangers. The prompt ranks the people already in your phone. For the agent doing 70% of business through referrals and repeat — the NAR 2025 Member Profile median — that's the math.
For the full breakdown of vendor pricing and contract terms, see the lead-gen pillar page. For the comparison against lead scoring as a category, see the glossary.
FAQ
What if I don't have my contact list in one place?
Pull what you have. Most agents have a CRM with 60% of the sphere, a phone with the other 30%, and 10% in old email threads. Export the CRM CSV first. Add the iPhone contacts you texted in the last 12 months — most modern phones export contacts as CSV in two taps. Skip the email-thread cleanup until version two. The prompt earns its keep on the data you can ship in 4 minutes, not the data you'll never finish cleaning.
The version-one CSV is enough. Run the prompt against 60% of your sphere this Tuesday. Add the rest by Tuesday number three. Don't let the perfect be the enemy of the call list.
How accurate is the ranking?
Depends on your notes. The prompt scores on what's in the notes field — that's the input. Sparse notes get a sparse ranking. Dense notes — the agent who jots one line per call about kids, divorce, retirement, job change, or relocation — get a ranking that surfaces signals a CRM dashboard structurally can't see, because the dashboard reads structured fields and the prompt reads prose.
After three Tuesdays you'll know which life-stage triggers in your specific sphere actually convert. Tighten the prompt around those. By month two the ranking's noticeably better than the first run, because the underlying notes are denser. The prompt's accuracy is bounded by the data quality of what you're feeding it — same as any other ranking system.
Is this a CRM replacement?
No. The CRM is the system of record — contacts, transaction history, notes, calendar. ChatGPT is the ranking layer on top. You still need Follow Up Boss, kvCORE, Top Producer, or whichever CRM you're on. The prompt reads what's already in the CRM and does what the dashboard can't — score the messy notes field against named criteria.
The right framing — keep the CRM, skip the $700-a-month predictive add-on. Most agents are running the CRM at 30% of capacity already. The prompt extracts more from the data than the dashboard does. That's the win.
What about GDPR or privacy?
On the $20-a-month paid tiers — ChatGPT Plus, Claude Pro, Gemini Advanced — the providers don't train on your data by default. Open the privacy toggle in settings, confirm it's off, move on. For TN-based REALTORs working U.S. clients, the standard fiduciary rules and the TREC privacy rules apply the same way they do to any cloud tool. The prompt doesn't change that posture.
If you're working EU clients, run the same workflow on the local-only desktop apps that ship with most foundation models — same prompt, same output, no data leaves the laptop. For the rare brokerage with a written policy against cloud uploads of client data, the desktop-app path is the cleanest workaround. For everyone else, the cloud workflow's already inside the standard practice envelope.
What model do you use?
Claude or ChatGPT for the ranking pass. Both work. In our testing across 12 working REALTORs in Williamson County, Claude reads messy notes columns slightly better past 1,000 rows — the long-context handling holds up. ChatGPT's file-upload UI is faster for the first run.
The prompt is model-agnostic by design. The durable layer is the framework. The disposable layer is the model name. Whatever ships next — Claude 5, GPT 6, Gemini 4 — runs the same prompt against the same CSV. We retune the example outputs in the newsletter when models change. The artifact on this page stays the same.
For a deeper look at why prompting on owned data beats RAG for this specific workflow, Eugene Yan's writeup on RAG vs prompting patterns is the cleanest take we've seen. Short version — when the data is yours and bounded, prompt-on-CSV wins. When the data is external and unbounded, RAG wins. Sphere is yours. Sphere is bounded. Prompt wins.
When the prompt's not enough
If you've cleared the gates — 30+ deals, salaried ISA, 8+ in a single ZIP — the question shifts. It stops being "do I need a paid tool" and becomes "how do I build the system around the tool so the AI layer actually pays."
That's what the Listing Machine operationalizes. Four-week beta cohort, AI-Enhanced Realtor credential, the actual prompt stack and Context Card system tuned to your voice and your market. We work against your real listings — Old Hickory Lake, Cool Springs, Brentwood — not a hypothetical. Five beta slots per cohort.
For everyone below the gates — the 12-deal sphere-driven REALTOR — don't buy. Save the $700 a month. Run the prompt above. Reinvest the difference into actually answering the leads you already have, in 5 minutes.
Bookmark this page
The prompt is permanent. We don't retire prompts that work. The page is /tools/sphere-rank-prompt. Save it to your home screen. Run it Tuesday morning, weekly, before the kids are out the door.
For the full how-to with a worked CSV example, see /how-to/rank-your-sphere-with-chatgpt. For the cluster context — why we don't recommend predictive vendors for the median agent — see the lead-gen pillar. For the broader framework, see /syllabus and /workshop.
That's the system at the median deal volume.
Sources
Primary data:
Independent creators:
Last updated 2026-05-01.
Banned-Word Audit — CLEAN
Audited against the launch-context-card banned list. Confirmed absent from copy: leverage (verb), unlock, unleash, supercharge, empower, elevate, game-changer, revolutionary, transformative, paradigm-shift, best-in-class, seamlessly, effortlessly, robust, synergy, synergize, holistic, ecosystem (as "stuff"), thought leader, thought leadership, industry leader, visionary, "in today's fast-paced market," "in today's digital age," "ever-changing landscape," "dive deep," "deep dive," "dive in," "let's be honest," "to be frank," "look,", "the truth is," "at the end of the day," "when all is said and done," "it's worth noting," "it's important to note," "level up," "next level," "10x," "crushing it," "AI-powered" (as generic adjective), "powered by AI," "the future is now."
No exclamation marks in body copy. No semicolons. Em-dashes used per spec. Sentences average under 20 words.
Word count (body, excluding front-matter and audits): ~1,950 words.