Workflow systems · Bangkok

Map the work first.Then build the system that runs it.

Most AI projects stall because nobody mapped how the work actually moves. We explore your workflows, design the system around them, and build it, with people deciding wherever judgement matters.

Past partners

  • Klook
  • ILCT
  • Fan & Fan Solicitors
  • CocoHarvest

From a client email to a filed matter, in one system.

This is a scripted replay with invented data. Pick an industry to see the same idea in a clinic or a travel operation.

saangai · Matter pipeline · Example Law Office

Step 1 of 7 · Email · 3 attachments

New request from a client

Example Hotel Co., Ltd. · HR

Please renew the work permits for our 14 foreign staff. Several expire next month. Passports and current permits are attached.

Attachments

  • work-permits-scan.pdf28 pages
  • passports.pdf14 pages
  • company-certificate.pdf3 pages

The work already has a shape. It lives in inboxes, LINE chats, spreadsheets and handoffs between people. We find that shape, remove the steps nobody needs, and automate the rest, so your team spends its time on the decisions only they can make.

How we work

Four phases. Each one leaves you with something you keep.

We start with how the work runs today, not with which model to use.

How we work
  1. 1

    Explore

    We sit with the people doing the work and trace one workflow end to end: every inbox, sheet, handoff and wait.

    About 2 weeks

  2. 2

    Design

    We redesign the workflow first, then the system: which steps disappear, which are automated, and where a person decides.

    About 2 weeks

  3. 3

    Build

    We build in short, demonstrable slices against your real documents and cases, so you see it working before it goes live.

    6 to 10 weeks

  4. 4

    Run

    We move it into daily use, watch it against real volume, and keep improving what the numbers show.

    Ongoing

What we build

The parts every system is made of.

  1. Workflow mapping

    /01

    How the work really runs today, drawn from the inboxes and sheets, not the org chart.

  2. Document pipelines

    /02

    Scans, PDFs and forms in Thai and English, read and turned into structured data.

  3. Agents with human review

    /03

    AI does the procedure. A person approves anything that needs judgement.

  4. Integrations

    /04

    LINE, email, calendars, sheets and the systems you already pay for.

  5. Private deployment

    /05

    Runs where your data is allowed to live, including on your own hardware.

Selected work

Systems in use and in delivery.

See all work

Law firm · Bangkok

In delivery

Royal Gazette reading and private legal search

Problem
Lawyers spent hours tracking new Gazette notices and searching their own past work by hand.
System
OCR over Gazette releases, a private library of the firm's documents, and cited answers that run on the firm's own infrastructure.

Meal-delivery operator

In use

Order routing and operations

Problem
Orders, driver routes, onboarding and payments were coordinated by hand across chats and sheets.
System
Routing, customer onboarding and QR payments brought into one operations flow.

Testimonials

In our clients’ words.

The new content auditing system, call operations system, and project management pipeline have kept us on the cutting edge of AI technology. With these innovations, we are able to function more efficiently, and compete within the competitive tourism industry in Thailand.

Klook

Travel and experiences platform

With this new LLM system, our firm will save hundreds of hours of billed legal work over the next year. Attorneys no longer have to manually look up files, and can easily use the system to review legal documents, manage cases, and prepare for trial.

Law firm

Thailand · Name withheld

Questions businesses ask first

What do you actually deliver?

A working system that runs part of your operation, plus the map and design it was built from. You own all three.

Do we have to replace our current software?

Usually not. We build around the tools you already use and connect them, replacing something only when it is the bottleneck.

Where does our data go?

Where you decide. We design for the data rules you work under, including PDPA and client confidentiality, and can deploy on infrastructure you control.

How does an engagement start?

With a short Explore phase: we map one workflow end to end and show you where the time goes. You decide what to build from there.

What happens when the AI gets something wrong?

Every step that needs judgement is routed to a person, and every automated step leaves a record you can check.