I build and run the operating systems behind small brands.

Sales reporting, order handling and ad accounts, plus training so your team can run AI on its own. From someone who has done the operations job, not an agency.

Book a callSee the work
1 day
to bring three sales channels and 8,964 orders into one P&L.
1/2
of one channel’s revenue was missing from the reports. The dashboard now reconciles it every hour.
−40%
cost per inquiry on paid search between July and September.
4.5 h
from brief to a live, indexed studio website.

Systems running in real businesses

Clients stay anonymous. Every number comes from the live system or its reports.

  1. 01

    Beauty brand with retail distribution

    Every sales channel in one hourly P&L, and the manual work around it automated.

    Also running: retail purchase orders handled from inbox to shipment on a server, a 902-product competitor review tracker, and 77 scheduled jobs.

    8,964
    orders unified on day one across three channels
    58%
    of warehouse product-linking was duplicate work. Removed, with 11 wrong links caught.
    10 / 16
    search campaigns were not tracking conversions at the agency audit
    ~25%
    of support inquiries filtered by an AI draft bot, at two to three cents a draft
  2. 02

    Hospitality venue in Southeast Asia

    Paid search rebuilt from zero tracked conversions.

    −40%
    cost per inquiry, July to September
    −46%
    blended cost per lead on Meta in three weeks
    5.8 → 4.0
    average Google position, with weekly reporting automated
  3. 03

    Boutique fitness studio

    Brief to live site in four and a half hours. The owner now edits it with an AI agent.

    4.5 h
    from brief to a live site
    31 min
    for domain, analytics and search registration

Four ways to work together

I take two clients at a time, so the person you talk to is the person who builds it.

  1. Operations Sprint

    2 to 4 weeks

    I map where your team loses hours, then automate the top three: dashboards, order and PO flows, inbox triage, reporting.

    • One source of truth for sales and margin
    • Automations running on a server, not a laptop
    • A written runbook your team can own

  2. Growth Systems

    Monthly

    I audit the ad accounts, fix tracking, then run a weekly loop of reporting and creative tests that someone actually reads.

    • Conversion tracking you can trust
    • A weekly report across ads, analytics and sales
    • Creative refreshed before fatigue sets in

  3. Launch Kit

    1 week

    A fast, multilingual site with analytics and search set up, handed over with a manual your team’s AI agent can follow.

    • Site, analytics, Search Console
    • Structured data and llms.txt
    • An agent-ready manual for edits

  4. AI Training and Consulting

    Workshop or monthly

    Hands-on sessions that teach your team to run AI agents on their own work, and advisory on where AI should and should not go in your operation.

    • Workshops built on your real tasks, not demos
    • An adoption plan ranked by hours saved
    • Rules for approvals, data and accounts

One operator, AI agents, and a signature on every outbound action

Each business runs in its own AI workspace. A browser agent handles logged-in tools, but only after APIs and plain requests have been ruled out. Every session is archived nightly, and what it taught is loaded into the next one.

Nothing leaves without a signature

Emails, orders and form submissions are prepared by agents, then held. Each runs only after a Touch ID signature bound to that exact content. Payments and passwords are never delegated.

Finished means checked live

Every deploy is checked on the live site and on mobile before it is reported as done. Silent failures are caught by exit codes, empty-result checks and heartbeats.

Mistakes are written down as rules

Each week the system turns its own sessions into lessons, and the next session starts with them loaded.

3,167
browser-agent sessions run
205
approval cards in 17 days, 44 of them rejected
257
lessons turned into working rules
20
background jobs running unattended

Also built

Travel publication
My own. 438 articles in four languages from one pipeline, with an image step and a writing-quality gate. Weekly Google clicks roughly doubled in two months.
Consumer AI app
Astrology, saju and Thai fortune in one product. Social login, a coin ledger and 579 backend tests. Shipped.
Supplier intelligence
A 288-page image-only trade-fair catalogue turned into 76 scored companies and 197 products.
Bilingual video pipeline
Eight product walkthroughs in English and Thai with AI voice, fitted music and burned-in subtitles.
Market-entry pack
A boutique hotel’s Korea plan: a nine-competitor teardown, with pricing checked live against five rivals on OTAs.
Showroom checkout and kiosk
A four-language payment plugin, and a kiosk that prints fortunes on receipt paper. In use since August.

I’m MK.

I work between Seoul and Bangkok. This year I have been running sales planning for a consumer brand while building the systems that run it, and shipping sites, ad systems and automations for hospitality, fitness and B2B marketplaces on the side.

I am not an agency. You work with one person who has done the operations job and can build the software for it.

Based in
Seoul and Bangkok
Works in
English and Korean
Capacity
Two clients at a time