Field-proven, first-person

Case studies

Not client work — my own businesses, automated first. That's the point: everything Callicarpa offers was proven in a real Texas brokerage before it was ever offered to anyone else.

Brokerage operations

I automated my own brokerage

“If AI can run the paperwork side of a Texas brokerage, it can run yours.”

The problem

A boutique DFW brokerage ran every transaction manually: contract data re-typed into timelines, milestone emails written from scratch, commission disbursements built in paid third-party software, market reports assembled by hand from MLS exports. The broker — also the owner, marketer, and bookkeeper — was spending evenings on paperwork instead of clients.

What was built

Executed contract PDF → extracted dates, terms, and amounts in one pass — no re-typing
Automatic transaction timelines, deadline checklists, and calendar entries
Milestone client emails (under contract → option ended → pre-closing) drafted on trigger phrases
Commission Disbursement Agreements plus a running transaction log — replaced the paid subscription entirely
MLS CSV → branded market report with pricing strategy
Multi-offer comparison tables for seller presentations

The result

Five to seven hours of admin eliminated from every transaction. Across a two-agent brokerage closing 25–40 transactions a year, that's roughly 125–280 hours returned annually — three to seven full work-weeks. One software subscription retired (comparable CDA and transaction-tracking services charge many times more), and every client email consistent and on time — regardless of how busy the week was. The entire system is documented and teachable, which is exactly what Callicarpa now builds for other businesses.

Document intelligence · Deep dive

Read the contract once, never re-type it again

“The contract gets read once — carefully — and then it's data.”

The problem

Every executed contract kicks off the same ritual: reading a 20-to-50-page package — contract, addenda, HOA addendum, seller's disclosure, proof of financing — and then re-typing the same names, dates, and dollar amounts into the timeline, the calendar, the client emails, the commission paperwork, and the books. The same field gets transcribed four or five times, and every transcription is a chance for an error. In a business where a missed option-period deadline costs real money, that's not busywork — it's risk.

What was built

Upload the full transaction package in any combination — AI reads every document in a single pass
Every field extracted into one structured transaction record: parties, dates, amounts, deadlines, commissions, financing terms, title, and survey
Each value is tagged with the document and section it came from, so discrepancies between documents surface immediately instead of at closing
Anything blank, unusual, or missing is flagged for review — never silently skipped
Every downstream automation — transaction timeline, calendar entries, milestone client emails, commission disbursement, close-out — runs off that single record; the PDFs never need to be opened again

The result

One careful reading replaces four or five hurried ones. Names, deadlines, and dollar amounts match across every email, calendar entry, and disbursement because they all come from the same extraction — transcription errors aren't just less likely, they're virtually eliminated — there's nothing left to re-type. And the pattern is the point: extract once, reuse everywhere works exactly the same for intake forms, invoices, and case files as it does for real estate contracts. It's the backbone of most Callicarpa engagements.

Seller representation · Deep dive

Offer weekend, without the 11 PM spreadsheet

“Sellers don't choose between contracts. They choose between outcomes — so show them outcomes.”

The problem

Offers arrive as dense legal PDFs — usually at night, usually on a weekend, and sometimes several at once. Comparing them fairly means extracting dozens of terms per offer (price, financing, earnest money, option terms, lease-back, contingencies, closing-cost asks), building a net sheet, and turning it all into something a seller can actually decide from — while every hour of delay puts the deal at risk.

What was built

Upload an offer PDF → every term extracted, with anything blank, unusual, or risky flagged
Seller net sheet: estimated proceeds after payoff, closing costs, and concessions — not just the headline price
Multiple offers → a side-by-side comparison ranked against what the seller actually cares about: price, closing date, lease-back, or certainty of closing
Response strategy plus drafted counter-offer terms and the exact language to send the buyer's agent
A branded, professional document the seller can read on their own time — consistent and decision-ready, every time

The result

What used to be one to two late-night hours per offer now takes minutes, and multi-offer situations — the highest-stakes, highest-pressure moments in a listing — become a clean side-by-side conversation instead of a pile of PDFs. Sellers decide faster and with more confidence, because they're comparing net outcomes, not paragraph numbers.

Product build

From idea to launched product: AmendReady

“Built by a broker, not a software company — that's why it works the way agents actually work.”

Visit AmendReady.com →

The problem

After a home inspection, buyer's agents face a 60–90 page report and a ticking option period. Turning that report into a repair amendment strategy takes hours — or expensive hourly help.

What was built

Upload an inspection report PDF → prioritized findings plus draft amendment language in minutes
Live Stripe payments with tiered pricing, automated results delivery by email
Client privacy scrubbing built in
Professional infrastructure — no dev team, no agency, one person and AI

The result

A live, revenue-ready SaaS product owned by Callicarpa LLC — designed, built, launched, and maintained by one person with AI. Every analysis saves an agent at least two hours of report review and amendment drafting. This is the proof behind every Callicarpa engagement: we don't sell theory; we ship.

Canva + AI · Listing marketing

One paste, every channel: the Listing Launch Kit

“Agents don't lose listings because their marketing is ugly — they lose momentum because it's endless. This makes it one step.”

The problem

A single listing needs a dozen pieces of marketing — Instagram post and story, Facebook post, LinkedIn content, Google Business Profile update, a property flyer for the sign rider, a print PDF — and then the status changes and every piece needs remaking: Coming Soon becomes New Listing, then Under Contract, Pending, Sold. Most agents make two or three pieces and give up on the rest.

What was built

Paste a listing address (or the data and photos from any listing) — AI pulls the details and writes the copy
A full graphics kit generated through Canva, built by a Canva specialist: Instagram post + story, Facebook page post, LinkedIn post + article header, Google Business Profile post, property flyer as graphic and print-ready PDF, and a Just Listed/Just Sold postcard
Every piece in all six statuses: Coming Soon → New Listing → For Sale → Under Contract → Pending → SOLD
Channel-matched text with every graphic: professional long-form for LinkedIn, short punchy captions with hashtags for Instagram, local-keyword copy for Google Business Profile
Status change? Regenerate the whole kit in one pass — nothing gets remade by hand

The result

The complete kit is 54 pieces — nine formats across all six statuses. Built by hand in a design tool at 10–15 minutes per piece, that's nine or more hours per listing; now it's a paste-and-review step. Every listing ships with the complete kit — every channel, every status — and the branding never drifts, because the templates are locked and the AI fills them.

Agent outreach · Deep dive

Personalized outreach at blast scale

“A blast gets deleted. A note about their buyer gets answered.”

The problem

MLS reverse prospecting tells a listing agent exactly which agents have buyers searching for a home like the one that just hit the market — the warmest cold list in real estate. But acting on it means writing dozens of individual emails, so most agents do one of two things: send a generic blast that gets deleted, or skip it entirely.

What was built

Paste the raw agent list straight from the MLS — any format; AI parses out names, emails, and buyer-match reference numbers
Automatic cleanup: duplicates removed, rows missing an email flagged, and a confirmation table shown before anything is created
One personalized draft per agent — greeted by first name, with their buyer-match reference woven into the subject line and opening so the agent knows this email is about their specific buyer
Drafts land in the email account ready to review — nothing is ever auto-sent; the broker reads and sends every one

The result

A pasted list of 30 agents — real lists have run as large as 250 — becomes one individually personalized draft per agent in minutes: outreach that reads like it was written one email at a time, because functionally it was. The drafts-not-sends design keeps the human relationship exactly where it belongs: with the broker. And the pattern generalizes — any business with a warm list can have personalized-at-scale outreach that a human still reviews and sends with one click.

Marketing automation

The AI marketing engine

“Consistency is the whole game in marketing. AI is how a small business affords it.”

The problem

Beyond the listing graphics, marketing a brokerage means constant writing and outreach: MLS remarks, agent outreach, client nurture, a social calendar — multiplied across every listing, competing with actual client work.

What was built

One property input → MLS-ready description plus four platform-specific social captions
Reverse-prospect campaigns: an agent list becomes one personalized email draft per agent, ready to review and send
AI virtual staging and photo enhancement, MLS-compliant
A repeatable Instagram content calendar

The result

Listing marketing that took two to three hours now takes about fifteen minutes per listing. Marketing went from a project to a routine — every listing gets the full treatment, not just the ones with spare time. Outreach that used to be a generic blast became personalized at scale.

Website build

A brokerage website that runs on data, not a webmaster

“The website stopped being a project and became a data file.”

See the live site →

The problem

Most agent websites come from a template vendor: a monthly fee, a layout shared with thousands of other agents, and stats that go stale because updating anything means a support ticket. The custom-agency alternative costs thousands up front — and still needs a developer every time a number changes. Either way, the site drifts out of date, because keeping it current is someone else's job.

What was built

A complete custom brokerage website — modern design, animation, mobile-ready — designed, built, and launched with AI in days, with no agency and no template subscription
Every fact on the site lives in one data file: closings, sales volume, reviews, communities served, active and sold listings — change a number once and every page, sentence, and stat updates with it
Portfolio photos pull straight from the brokerage's existing listing tool — nothing re-uploaded, nothing maintained twice
A stylized North Texas service-area map in the hero instead of the stock agent-photo layout every other real estate site uses
Search-engine plumbing built in: structured data, sitemap, per-page titles and descriptions
Free hosting with automatic deploys — push an update and the live site refreshes in minutes

The result

A live brokerage site with a verified track record — 419 closed transactions, over $130 million in volume, 101 Google reviews at 4.9 stars — maintained by one person who edits data, not code. A new listing is one entry in one file; the site rearranges itself around it. And it does double duty: it's the brokerage's working storefront and the live demo Callicarpa shows prospects — proof that a small business can own its entire web presence and keep it current without a vendor in the loop.

Get your free AI audit