Messy in · Names masked first · AI reads, understands, checks, generates · Clean outMove your mouse to pluck the ink line
Into the grainData platforms and business systems

Know the business first.
Then build the system.

INGRAIN builds data platforms and business systems for organizations and companies.

From website leads and sales follow-ups to orders and production, HR and finance, and class scheduling: whatever still runs on spreadsheets and group chats, we turn into a system built the way you work, with AI wherever it helps.

  • You own the source code by default
  • On your own servers whenever possible
  • Working prototype, usually in a week
  • Systems in use in mainland China, Japan, and Hong Kong
Industries served
  • Design and construction
  • Services
  • Printing and manufacturing
  • Education and training
  • K–12 education
  • Customer service
  • Telecom compliance
  • Brands and websites
Work

Live systems do the talking.

All real projects, live, in pilot, or in acceptance, with client details removed. Interfaces are redrawn from the real features, with fictional data. They play on their own, and some parts are clickable. Below each case is its business workflow; steps marked AI use AI.

ExamplesThree real workflows: how they ran, and the systems they became
LiveDemo data
Before · Sales chatFollow-ups buried in messages
ESEast Region Sales (8)
K
Kevin · 10:02
@Sarah Any word from Grandview Materials? The boss is asking about the contract.
S
Sarah Chen · 18:40
AI
Sales assistant · 18:40
Split into 3 records. Matched 2 existing clients. Confirm to save.
After · Sales & projectsSarah Chen · Today
Collected this month¥96,500
New follow-ups today0
Needs your review0
Follow-upsAuto-generated from chat
EI
Eastline Interiors · Design revision

Mr. Liu · New version due by next Tue

To-do · Tue
¥
Eastline Interiors · Contract ¥320,000

2 installments · First ¥160,000

Payment plan set
GM
Grandview Materials · Follow-up

Early next month · Sarah reminded

Reminder · Next month
Payments above the contract amount are blocked automatically✓ Confirmed and saved
Before · Excel back and forthHR ⇄ Payroll provider
▦Nov_payroll_FINAL_v3(2).xlsx
NameBaseOvertimeInsuranceNet
Tanaka280,00012,40041,200251,200
Li260,0008,0OO38,300#VALUE!
Sato310,000045,600264,400
Wang240,00015,20035,400219,800
Yamamoto295,000?43,500251,500
Chen255,0006,30037,600223,700
Suzuki270,0009,80039,800240,000
The provider’s version doesn’t match ours. Which cell??
After · PayrollNov · 3 companies
01HRsubmits
02Send toprovider
03Importand check
04Fill inand confirm
05Finalize
06Export
EmployeeNetProviderCheck
Ichiro Tanaka——Match
Na Li——Match
Ken Sato——Match
Lei Wang——Match
Hana Yamamoto——Match
payroll.pdf bank_transfers.csv One-click export
Before · Paper sign-inSign-ups over WeChat DMs
Bookkeeping 101 · Sign-in sheet 10/30
  1. Xiaoming Wang ✓
  2. Fang Li ✓
  3. Ken Sato Leave
  4. Wei Zhang Wei Zhang Late
  5. Jing Chen ?
  6. Hanako Yamada ✓
After · Roll call on the teacher’s phoneStudent, teacher, and admin apps
Thu 19:00 · Room 302
Bookkeeping 101 · Lesson notes
Xiaoming Wang
PresentLeaveLateAbsent
Fang Li
PresentLeaveLateAbsent
Ken Sato
PresentLeaveLateAbsent
Wei Zhang
PresentLeaveLateAbsent
Jing Chen
PresentLeaveLateAbsent
Present 0Leave 0Late 0Absent 0
This week’s classes · SeatsAdmin
Bookkeeping 10112/12 Full
Business Japanese N29/12
Business Writing4/10
Student app · RescheduleMobile
Ken Sato → same class 11/06Approved
Roll call synced to admin. No more paper attendance.
The three workflows come from cases 01, 02, and 06 below: an AI sales assistant on WeCom, multi-company payroll, and three-app roll call for an education provider.UI redrawn in code · Demo data
Case01Design-build firm, mainland ChinaDesign and build · 6 systems · Since May 2026In useEngagement · Phased

One company’s whole business, from website leads to collected payments, connected in 5 months.

Before: the website didn’t show up in search. Sales follow-ups were scattered across WeCom chats. Contracts and payments lived in shared spreadsheets that never matched. Designers each found their own AI tools, with scattered accounts and unclear costs.

  • Baidu and AI search now find and recommend the company, and inquiries come in through the website
  • The whole team uses the CRM every day: payments over the contract amount are blocked automatically, commissions calculate themselves, and AI writes a morning briefing
  • AI drafts three RedNote posts a day and one gets picked; data from 7 platforms comes back daily
LeadsWebsite and CMSRedNote content pipelineWeChat Channels and social leads
SalesWeCom AI sales assistantSales and project management
Design and deliveryIntranet AI renderings
See this case
CRM dashboard: customer allocation, team follow-ups, monthly report
All-platform dashboard
Intranet AI renderings
Case02Service company, JapanServices · 4 systems · Since May 2026In useEngagement · Phased

It started with one attendance feature. Six months later, it’s the company’s internal platform.

Before: how many days foreign employees had lived in Japan, and how long their visas had left, were counted by hand, and one mistake could jeopardize a renewal. Application forms were copied from passports letter by letter. Payroll for several group companies went back and forth in Excel with an outside provider.

  • Payroll live in 6 days: the provider’s figures are checked automatically, and payroll sheets and bank transfer files export in one click
  • Daily report review live in 4 days: the system lines up whom managers should read first; AI only ever sees code names
  • ID documents never leave the company; forms fill themselves, and people only check them
Onboarding and attendanceAttendance and visa tracking
Visa applicationsID scan to formMail management
Daily operationsPayroll reconciliation with providerDaily report review
See this case
Payroll: six steps, mismatched rows highlighted automatically
Mail log
Case03Printing company, Hong KongPrinting and online retail · 2 systems · 6 months in operationIn useEngagement · Fixed price

Customers order on their phones, layout happens automatically, and paid orders came in on day 5.

Before: orders came in over WhatsApp, a dozen or more messages each. Designers laid out every card by hand and couldn’t keep up in peak season. Fitting several cards onto one sheet meant lining them up by hand, and every misprinted sheet was money lost.

  • Day 5 after launch: first real paid orders; running for 6 months since
  • Five languages, one mobile ordering flow, Traditional Chinese by default
  • The Mac layout app matches the physical mold dimensions, flags images that aren’t sharp enough, and uses AI to cut out the white underlayer
OrderingMobile order wizard
ProductionProduction adminMac layout app
See this case
Production admin: 5 cards per A5 sheet, landscape auto-rotated
Order wizard
Mac layout app
Case04Corporate websitesService line · 3 systems · Since 2026In useEngagement · Fixed price

Websites for a whole group of companies, designed, launched and managed on one platform.

Before: each company hunted for its own outsourcer, and timelines and styles got out of hand. Once built, no one kept the sites updated or answered inquiries. Content sites wanted to be found by search engines and AI search, with no proven way to get there.

  • 18 websites, all live in 5 days, each with its own design, picked from 40+ original concepts
  • Chinese–Japanese websites for 16 companies, 178 pages through internal review
  • A multilingual content site template, running an eight-language guide site with 690+ articles
DesignA custom design per site
LaunchMulti-site adminChinese–Japanese sites at scale
ContentMultilingual content site template
See this case
Multi-site console: site list, review and launch progress
Case05AI employeesBuilt in-house · Content ops, video, hiringIn use

Publishing, video production, resume screening. Handed to AI employees.

Content has to go out every day, comments need replies, numbers need checking. Videos get made one at a time. When hiring, resumes pile up and each one needs reading. The work is repetitive and never stops, and it all used to land on a few people.

  • AI operations: publishes on schedule and replies to comments across major content platforms, mostly fully automated, a few semi-automated; reviews its own day and writes a daily report. In use at many companies
  • Short-video production farm: ads, short dramas, music videos and series go from script to final cut automatically; if a shot isn’t right, only that shot is redone
  • AI recruiting assistant: screens resumes against the role, reviews portfolios, drafts a reply for every candidate, and compiles qualified ones into an evaluation sheet
OperationsAI operations
VideoShort-video production farm
HiringAI recruiting assistant
See this case
What we solve

Does this happen
every day?

This is what most companies and organizations tell us when they first reach out. We find the one that hurts most, and start there.

“Getting one number means waiting for several people to merge spreadsheets.”

  • Everyone reports a different number. No one knows which to trust.
  • You learn the month lost money only when it ends, too late to adjust.
  • Half of every meeting goes to reconciling numbers instead of making decisions.
What we do

Data is entered once, in one place. Reports and dashboards build themselves, and every morning the numbers you need land on your phone.

Related casesCase 01

“When a veteran leaves, the clients and the know-how leave too.”

  • Client details and chat history live on employees’ personal phones.
  • Only one person knows a key step. When they’re out, everything stops.
  • New hires need a month of hand-holding from senior staff.
What we do

Clients, records, and files stay in the company’s system. The steps are written into the system too, so new hires can follow them and get up to speed.

Related casesCase 08Case 01

“Our staff spend half their time copying sheets, reconciling accounts, and pasting.”

  • The same information gets entered again in every spreadsheet, chat, and form.
  • Text on IDs and invoices is typed in by hand, one character at a time.
  • Month-end reconciliation means days of overtime for finance.
What we do

Automate what can be automated: invoices and IDs are read and filled in for you, data on both sides is matched automatically, and only the mismatches go to a person.

Related casesCase 02Case 03

“We only find problems after the fact, and can’t tell which step went wrong.”

  • IDs and contracts are about to expire, and no one says a word.
  • Late payments and missed shipments surface only when the client calls.
  • Numbers in the sheet get changed, and no one knows who did it or when.
What we do

The system watches for expirations, overdue items, and anomalies, and alerts the right person ahead of time. Every edit is logged, so you can see who changed what.

Related casesCase 02Case 01Case 08

“A client asks for a status update, and we have to ask around before we can answer.”

  • Slow replies send prospects to a competitor.
  • Follow-up depends on each salesperson. When a deal is lost, no one knows why.
  • Placing one order takes over a dozen messages back and forth. Clients find it a hassle.
What we do

Clients order and book on their phones, without waiting for a reply. Follow-ups are logged automatically, and the system tells you who to contact today.

Related casesCase 03Case 01Case 06

“The software we bought didn’t fit how we work, so we went back to Excel.”

  • It was built around someone else’s process, so we had to bend to it.
  • Our systems don’t talk to each other, so data gets moved by hand.
  • Once it shipped, the vendor was gone. Changing a single field meant a new quote.
What we do

We sit beside your team to see how the work really gets done, then build around your process. We connect to the systems you already have wherever we can. After launch, monthly support is available, and you can raise requests anytime.

Related casesCase 02

“Everyone says AI boosts productivity. We don’t know where to use it, and we’re afraid of leaking data.”

  • Everyone uses their own AI tools: scattered accounts, messy costs, no real process.
  • Client and employee data feels too risky to hand to outside AI.
  • When AI gets it wrong, no one catches it.
What we do

AI goes only where it truly saves work. Before anything reaches an outside AI, names of people and companies are replaced with codes. AI output is saved only after a person confirms it.

Related casesCase 01Case 02

Not on the list?
We build that too.

These are just the most common cases. We don’t sell off-the-shelf software. Every business system is built from scratch around the client’s own process. However niche the industry, however unusual the process, even if the devices or platforms you need to connect have never been integrated before, we can build it.

New industry? We go on site and learn the business before we build. That’s what we mean by “Know the business first. Then build the system.”

Tell us about your situation
Four things we do most
01

Business systems built around your process

Data, workflows, permissions, and roles are shaped by your business, replacing spreadsheets, no-code tables, and manual logs.

02

AI for the business you already run

AI reads invoices and images, understands conversations, checks data, and generates content, plugged into your existing workflows. AI output is saved only after a person confirms it. Models can be swapped for Chinese ones on request, or run on your own servers. Where AI doesn’t help, we leave it out.

03

Web, desktop, and mobile, delivered together

Web apps, Mac / Windows desktop software, mobile, WeCom / Feishu bots: one team builds for every platform. We also build websites and multi-site setups.

04

Multilingual, multi-region

Interfaces, websites, and documentation can ship in multiple languages. Local rules in each country and region, such as IDs, social insurance, and payments, are built in to local requirements.

Business firstWhy this order

Before the first cut,
a carpenter snaps an ink line.。

Before we write a line of code, we get clear on the business: who fills in the forms, who waits on the results, which step breaks most often. Snap the line true, and the system comes out straight. What we build, the team wants to use the very next day.

INGRAIN / dyed into the grain · 入木三分
How we work

From the first conversation,
four steps to launch and beyond.

Most outsourcing vendors handle only the middle two steps: take requirements, write code. We learn the business before, and stay on after. One team from start to finish, with one lead as your contact throughout.

Two hands snapping a blue ink line across a light wooden board, with an ink pot and marking gauge beside them
One layer in
01

On-site discovery

We sit beside your team and watch how they fill in forms, track orders, and reconcile accounts every day. Then we rank the pain points by impact.

1 hourdiscovery call, then 1 to 3 days of research, mostly remote, or on site with you
The first blank sawn along the ink line, next to a Japanese saw
Two layers in
02

One-week prototype

A first version on real data that people can click through and use. Your team tries it, and we refine it as they go.

1 weekOn your real data. Not happy? You can stop there.
Two pieces of wood joined tight with a dovetail, ink lines aligned
Three layers in
03

Launch and handover

Installed on your own servers whenever possible. We migrate, clean up, and verify your historical data, set up backups and permissions, and hand over the user manuals.

1 to 4 weeksDepending on system size, signed off stage by stage
Oiling the finished joinery with a cotton cloth
Into the grain
04

Ongoing improvement

Every request gets a delivery log entry. The system grows with your business instead of ending at handover.

MonthlyPriced by system size, quoted when we discuss requirements
Engagement models

Choose what fits, or combine them.

We give a price range once we understand your needs. Formal contracts, invoices, and NDAs are all available, with itemized quotes.

01

Phased

Start with one starter system. Once it runs smoothly, grow outward one system at a time, each quoted separately.

For teams that want to try one piece first, then expand
02

Fixed price

Scope and total price are set before signing. You pay the balance at acceptance, with no extra charges within the agreed scope.

For projects with clear requirements and fixed scope
03

Service fee + profit share

A lower service fee up front, then an agreed share of revenue once the system starts earning. We can also turn the system into an industry product together and share the revenue from selling it to your peers.

For companies with a tight budget but a system that can bring in revenue, or that want to turn their way of working into an industry product
04

License

No source-code buyout. Pay an annual license fee for a lower up-front cost, and we handle upgrades and maintenance.

For teams that want to spend less and don’t need to maintain code themselves
Delivery and commitments

Everything handed over.
And we stay on.

Every project includes all of this by default. Nothing extra to negotiate.

What you get

Delivery packageSample project · v1.0Packing 0/6
  • Full source codeYours by default, except under a license
    Pending
  • One-click deploy and rollbackRoll back to the last version in one click if something breaks; a new team can take over
    Pending
  • Installation and permissionsOn your own servers whenever possible; the system controls who sees what
    Pending
  • User manualsA technical edition for engineers and an operations edition for staff, multilingual on request
    Pending
  • Delivery logEvery request traceable from ask to launch
    Pending
  • Daily automatic backupsWith step-by-step restore drills
    Pending
INGRAINYour server

Four worries, handled by default

StartBook a call

Bring what you need.
Let’s talk for an hour.

Not sure how to describe it? That’s fine. Bring the spreadsheets, chat screenshots, and paper forms you use today. We’ll look at how you work now and tell you where to start and roughly how long it will take. If it feels right after the call, we’ll follow up with a written proposal and a price range.

WeChat
LanguagesChinese · Japanese · English
FormatOnline, or on site with you
Into the grain