Skip to content
한국어

EZWorks · United States

Groundwork before AI

Whether AI can be attached is decided by where the state of the work lives — not by the tool. Inside a system, or in someone’s head and a spreadsheet.

Sixteen years inside the US subsidiaries of Korean companies — working the gap between a head-office system and how the work actually runs locally.

01

Problem

One click builds anything, the story goes. Applied to your own company it comes apart. The cause is not the model — it is that the work was never defined.

02

Order

Define the process, fix its inputs and outputs, let data accumulate, then AI. Each step is the precondition for the next. It is not a menu.

03

Start

A two-week review at a fixed price. It produces a read on the current state, three bottlenecks, and a do-now / do-later / do-not list. Those remain even if you stop there.

The Reality

A structure where people become the bottleneck

Work cannot be divided the way a large company divides it, so the owner and a few early people carry effectively all of it.

Your own work starts after hours

Customers and incoming requests consume the day.

Procedure lives in people, not documents

Time off is a burden. A resignation is an incident.

Nothing records where the work stalls

Always busy, with no way to tell what to fix.

The same report is rebuilt every month

By the time it is finished the numbers are already old.

Where To Look

Where to start reading

There is an order to these. Either way they arrive at the same place.

Services — four areas, with the work in each

Assessment and definition, operational systems, integration and visibility, automatic decisions. Each area carries the work actually done in it.

See the services

How it runs — segments and exits

Assessment, definition, build. The end of each of the first two is a formal exit. The nine deliverables and the fixed review price are here too.

See how it runs

Products — what the procedure produced

Built while applying the same procedure to our own work first. They also serve as the first thing adopted after an assessment.

See the products

First Application

The first thing it was applied to

The order set out above — define the process, fix inputs and outputs, accumulate data, then AI — was first applied not to a client but to our own operation.

Starting a business brought all of it at once: expenses, invoices, receivables, turning what was said in a meeting into requirements, keeping track of what still had to be done. Memory did not scale. So the same procedure was applied to us first.

Our own workExpenses · invoices · receivablesRequirements from meetingsTo-do · state of playHeld together by memoryThe same four stepsdefine → in/out → data → AIidentical to what we proposeWhat had to be built along the wayVoice records — so nothing said is lostExpenses · approval — for what returns monthlyIntake · tracking — to give requests a numberThis is where the procedure is shown to workThe same procedure is applied to client work. The products are evidence of what it produces.

These were not built in order to be sold. They are what came out of the same procedure.

So the products do two things. One is to serve as evidence that the procedure actually produces working systems. The other is to be a finished starting point for the first thing adopted after an assessment. A company whose requests are scattered across chat and email does not need intake and tracking built from scratch.

The scope shows here too. Not only web screens — mobile applications and store release sit inside the same procedure. Where an app is what is needed, an app is what gets built.

Product or internal system, none of it began from a requirements document handed over. Each started by finding the problem and deriving the requirements first-hand.

Products

What we build and run ourselves

These came out of applying the procedure above to our own work. They attach to what you already run — SAP, Microsoft 365 and the Power Platform, accounting software including QuickBooks, and CRMs.

EZClass

Members, attendance and billing in one place.

An operations system for multi-site academies and studios. Billing conditions that differ per member are held as rules, so nothing is recalculated by hand each month.

  • Billing becomes lookup
  • No payment method, no auto-run
  • Per site and consolidated

Get in touch about EZClass

EZHaru

Record. Everything else is done for you.

An AI voice recorder that files itself. Tap a category and record — the transcript, the summary and a tidy note file appear in your own cloud, already in the right folder.

  • Files itself
  • Transcribes and summarizes automatically
  • Learns your vocabulary

Explore EZHaru

EZWorks Expense

Receipts in. Bookkeeping done.

A receipt and expense workflow for companies that still move numbers by hand: capture, read, route for approval, and hand off to accounting.

  • Capture where the receipt happens
  • Reads the document for you
  • Approval that follows your rules

Get in touch about Expense

Summary

In short

  1. 1AI does not fit your company because nothing records where the work starts and ends — not because the model is weak.
  2. 2Define the process, fix inputs and outputs, accumulate data, then AI. In order it takes weeks; in reverse you are back at the start half a year later.
  3. 3Two things set the starting point — whether the process is defined, and whether core data lives in a system of its own.

One thing next

Name the process that takes the most hands, in one line. That line sets the scope of the review.

Get in touch — one question

We answer first whether this is a situation that warrants a review, or whether it is not the moment.