The AI team inside your business.

We run the work that moves your numbers, own the result, and know your business better every month.

Book a call See the builds
Today's list19 houses pricedrunning
real figures, streets hidden6 Oct run

Tools we have built with

  • Shopify
  • WhatsApp
  • Instagram
  • Google Drive
  • Google Sheets
  • Gmail
  • Telegram
  • Slack
  • HighLevel
  • Notion
  • Airtable
  • YouTube
  • Metricool
  • Spur
  • Stripe
  • Calendly
  • Claude
  • ChatGPT
  • n8n
  • Supabase
  • Vercel
  • Trigger.dev
  • Remotion

Five businesses. What changed for each.

$4K+a month in content savings
Ji Martial Arts

Videos made and shoppers answered, without anyone doing it by hand.

E-commerce: Ji Martial Arts, an Indian fashion brand

~15minto price a deal, down from about a month
A Dallas real estate investor

A priced list of houses every morning, and deals priced in about 15 minutes, not a month.

Real estate: a Dallas real estate investor

We run our own day on three tools we built.

Our own tools

3rdof 200+ teams
Swans Applied AI Hackathon

A legal intake that reads police reports, and a case timeline for cents.

Hackathon

A US AI agency, under NDA

Analytics, social, lead capture, portals and checkout for its construction and real estate clients.

Systems that keep running and get better every turn.

What we have delivered

  1. A recruiting agency2 minper candidate intake, from 45
  2. Swans Applied AI Hackathon 3rdof 200+ teams
  3. A Swiss software company~25 hoursa month back (estimated)
  4. A Dallas real estate investor~15 minto price a deal, down from about a month
  5. Ji Martial Arts$4K+a month in content savings
  6. An Indian fashion brandRs 1 lakh+in orders from WhatsApp messages in one sale
  7. A client in IndiaWhatsApp AI image stylistin progress
  8. Exhibit A HackathonCentsa case timeline, against $28 to $150 at market rates
  9. A US AI agency, under NDAAnalytics and attribution, organic social, lead capture, client portals, funnels and checkout

E-commerce

A martial arts brand gets its videos made for it. A fashion brand answers its shoppers on its own.

$4K+a month in content savings Ji Martial Arts
Ji Martial Arts, a martial arts e-commerce brand

Workout videos made and scheduled from one message.

A workout video maker. Ask for a 12, 15 or 40 minute workout in a chat. It makes the video, saves it to Drive and schedules it.

An Indian fashion brand

Shoppers who leave a cart or comment on a reel get a reply with the link, on their own.

Rs 1 lakh+in orders from WhatsApp messages in one sale

A consultancy for the brand: ads, Shopify and the backend work, plus the automatic WhatsApp and Instagram replies.

Real estate

Houses behind on taxes arrive already priced, and a deal gets its offer in about 15 minutes.

~15minto price a deal, down from about a month A Dallas real estate investor
Today's list, most behind on taxes first

A priced list of houses behind on taxes, every morning, with nobody reading county sites.

Our own product, made for house buyers. It reads the county tax roll every day, finds recent sales nearby and works out the most to pay for each house. From the 6 Oct run, streets hidden.

The third house, opened: Grand Prairie, TX
Behind on taxes
$30,236
owed to the county
Fixed-up value
$280,139
from 4 recent sales
Most you should pay
$121,158
after repairs and fee
Room after taxes
$90,922
most to pay, minus owed

$280,139 x 0.7, minus $66,990 repairs, minus $7,949 fee = $121,158

A Dallas real estate investor

From about a month to about 15 minutes.

When a deal moves to Underwrite in his CRM, the engine picks recent sales nearby, works out the most to offer and writes the answer back.

Hackathon

3rd of 200+ teams at the Swans Applied AI Hackathon, and a second build for its next edition.

March 2026: a personal injury law firm, as set in the challenge

About 90% less data entry (estimated).

Upload a police report and the AI reads out the details. A paralegal checks them, then the case is created and the retainer is emailed.

July 2026: Exhibit A, for case chronologies

A case timeline for cents, against $28 to $150 at market rates.

It reads a case's treatment records into one timeline and draws the gaps in treatment that the other side will attack.

AI agency, under NDA

We build for an established US AI agency serving construction and real estate firms.

  • Analytics and attribution
  • Organic social
  • Lead capture
  • Client portals
  • Funnels and checkout

Our own tools

We run AugmentLoop on six tools we built ourselves.

Pinch to zoom, drag to move
Pipeline Viewer showing our underwriting engine as a diagram of connected steps, with the list of other systems down the left.

Pipeline ViewerEvery automation we run, drawn as a diagram you can click through. A real screen with the names taken out.

The planner's Today screen: the day's tasks, habits and what needs attention, in a light layout.

The plannerOur whole day on one screen: tasks, habits and what needs us. Shown with sample data.

The phone agentMessage it on Telegram and it saves the note. Each morning it links one old note to a reel we already saved. Sample chat, invented task.

The reel library: saved reels as small cards grouped into topic clusters on one canvas.

The reel libraryMore than 1,250 saved reels on one canvas, sorted by topic. Each one gets a one-line takeaway we can ask for by number.

Where the timeline above came from

Reel 42The active step lights, then ticks, and the wire draws on to the next one.
Reel 1196A depth carousel: the centre card full size, side cards smaller and tilted back, all stepping together.
This site's own timeline: a line of results with the first three lit and a signal running along it to the next.
This site's timeline.

We pulled both reels out of the library by number. Reel 42 became the lit step and the line that draws to the next result. Reel 1196 became the carousel the timeline turns into on a small screen.

Claude Code Mission Control: bar charts of spend per day, in blue and orange, with a legend of who spent it.

Claude Code Mission ControlA meter for what our AI coding costs, per day, per person and per model.

The Hackathon Radar board: a daily list of hackathons, each with prize, deadline and a fit score.

Hackathon RadarA daily sweep of hackathons, scored for how well they fit us. This is the board as published.

The lead engine, in its own light look

The lead engine's own dashboard, as built: the ranked list, the map, the radar and a deal card. Invented Dallas data, owner names hidden. Map tiles © OpenStreetMap contributors.

The lead engine's ranked list: each lead with its score, back taxes and an offer range, owner names hidden.
The list
The lead engine's map: a ranked list beside a map of Dallas with a pin for each lead.
The map
The lead engine's radar: coloured bands over Dallas showing where the most unpaid tax sits, with the top leads listed beside it.
The radar
A deal card: one lead's price, what it owes and what to offer.
A deal card

More we have built.

A recruiting agency

2min per candidate intake, from 45

Candidate intake went from 45 minutes to 2. About 15 hours a week back for the team.

Forms feed straight into the recruiting CRM, with no duplicates and nothing dropped.

A Swiss software company

~25hours a month back (estimated)

Invoices filed without anyone touching them.

Invoices move from the ERP into the document archive, filed and indexed.

A client in India

In progress

A WhatsApp AI image stylist.

Ready to stop doing ops by hand?

No pitch deck. Walk me through what's eating your team's time, and I'll tell you if automation makes sense. If it doesn't, I'll say so.

Book a call

A 30-minute call.