Macro and Automation Program Cost: What Workflow Automation Actually Prices At [2026]

The median automation quote Freesi (freesi.net) issued in 2026 was 400,000 KRW (approx. USD 290). That is a sample of five — too small to call a market average, and biased toward the small one-off automation requests that tend to reach us. Those limits stated, the number stands as it is.

The range we actually see in practice is wider than that. Work that begins and ends inside an Excel file starts around 150,000 KRW (approx. USD 110); handing over an entire flow that crosses several programs passes 1,000,000 KRW (approx. USD 725). Same phrase — "one macro" — six times the price.

That spread is exactly why quotes stop feeling trustworthy. So rather than opening with a price table, this starts with why one word produces so many different prices.

Saying which of these four you mean makes the reply you get back dramatically more accurate.

1) Excel macro (VBA) — work that begins and ends inside a spreadsheet: cleaning up sheets, normalizing formats, merging files, printing into a fixed template. It never leaves Excel, so it is the cheapest tier.

2) Screen-click automation — a script operates an existing program's screens in place of a person: keying values into an in-house ERP, clicking through a website in a fixed order. This is the technique you use when the target system offers no API. Its weakness is that it stops working when the screen changes, which is why a maintenance conversation always comes attached.

3) Workflow automation (RPA) — the whole chain across multiple systems: pull an attachment out of a mailbox, tidy it in Excel, upload it to an internal system, post a summary to a chat channel. Judgment and exception handling sit in the middle, and that is what raises the cost.

4) Purpose-built software — not automation but a redesign of the work itself: retire the spreadsheet, build a database and screens instead. From here on it is software development, not a macro.

Most of the disappointment in this market comes from being quoted at tier 4 for a job that ends at tier 1, or from expecting tier 1 pricing for a job that is really tier 3.

One-off build, excluding Korea's 10% VAT.

Two items stack on top of these. Running unattended on a schedule requires a machine for it to run on (see server and cloud costs), and a screen where a person reviews the results adds its own build. If you start it yourself by pressing a button on your own PC, you need neither.

If two vendors hear the same requirement and come back twice apart, usually only one of them counted one of these four.

Login and security steps. Whether an ID and password is all it takes, or a CAPTCHA or phone verification sits in the path, changes the difficulty completely. If a human has to clear the verification every run, the point of automating shrinks — so this is something to settle before quoting, not after.

How often the target screen changes. An in-house system we control is stable; an external site has to be repaired every time the other side redesigns. In those cases the maintenance terms matter more than the build price.

How far exceptions are handled. Automation that covers only the normal case costs about half. Defining what happens when a value is missing, duplicated, or malformed noticeably increases the work. But automation without exception handling ends with a person re-checking the output anyway — you spend the saved money back in time.

Volume. Ten runs a day and ten thousand runs a day are not the same code. Stating the volume up front reduces both over-quoting and under-quoting.

Plenty of people now build simple macros themselves with AI tools, and for tier 1 — work that ends inside Excel — that is genuinely worth trying.

Use this threshold: when hours saved per month × what your hour is worth × 12 months exceeds the build cost, hiring wins. A 30-minute daily task is 10 hours a month; at 20,000 KRW (approx. USD 15) an hour that is 2,400,000 KRW (approx. USD 1,740) a year. Against a 400,000 KRW (approx. USD 290) automation, it pays for itself in two months.

There are also clear cases where you should not hire at all.

The process is about to change — building after the change is cheaper

A task you do a few times a month — a checklist beats automation

The feature may already exist — check the automation menu inside the tool you already pay for

A vendor who asks "can your existing tools already do this?" before quoting is the more trustworthy one.

The steps you take today: which screen, what you click, where the result goes — step by step

Frequency and volume: how many times a day, how many records per run

The programs involved: Excel, an internal system, or an external site (name it)

Login and verification: whether a CAPTCHA or phone verification is in the path

When it runs: you press a button, or it has to run on its own at a set time

With those five lines, the first reply from any vendor comes back with numbers in it. For rates on adjacent project types, see data collection program cost and analysis program cost. For clients outside Korea, this kind of work runs fine fully async — requirements over chat, quotes in USD.