Descarty
日本語Download overviewRequest a demo
Menu
Request a 15-minute demoDownload the overview日本語

What RPA costs to keep running, beyond licenses

Author: , Founder, Descarty, Inc.Published: Last updated:

Ask about the cost of RPA and license fees come to mind first. Yet at companies that have run bots for a few years, the costs that weigh most never appear on a license invoice: the time spent fixing bots that stopped, the manual work until they are fixed, and the waves of rework when a browser or scripting language reaches end of support. These costs are scattered across departments’ working hours and are almost never added up.

This article sets out what goes into the cost of keeping RPA running, and shows how to total it with a worked example. It also summarizes, with sources, recent changes in pricing and support at the main RPA products.

What running costs include

Cost What it covers Easy to miss
Licenses Development and runtime licenses, management servers Usually annual, and exposed to price revisions and licensing changes
Environment PCs, servers, virtual desktops, and accounts for bots Every OS or browser update needs testing
Building Staff time, outsourced development, training Few people can build bots, and their time is taken
Fixing Handling screen changes and errors, outside maintenance contracts Scattered across departments’ hours and never totaled
Downtime Delays until someone notices, manual work as a fallback A stop on a closing date leads to overtime or late payments
Management Updating ledgers and procedures, inventories, audits Not kept up to date, then useless when needed
Rework and migration End of support for browsers and scripting languages, product upgrades Every few years, rework that touches every bot

Licenses and environment show up on invoices and asset registers. Fixing, downtime, and management are hard to see, and they are almost entirely people’s time.

Recent changes in pricing and support (as of October 2026)

In recent years, the main RPA products have kept changing their prices, licensing, and the support of the technology they rely on. If a renewal is near, check the following.

WinActor

  • Price revision: from January 1, 2025, prices of all WinActor products rose by roughly 10%. The annual license fee for a node-locked license is ¥1,098,680 for the full version and ¥300,080 for the runtime version (both including tax)
  • AI integration license: a separate license for generative AI features went on sale in October 2025. It currently costs ¥2,197,800 a year for the full version and ¥600,600 a year for the runtime version (both including tax). From October 2026, generative AI usage is capped, and usage above the cap requires buying additional packs
  • VBScript: Microsoft announced in October 2023 that VBScript will be retired, and plans to turn it off by default around 2027. WinActor has no plans to add to or fix its VBScript libraries, and the feature that helps convert VBScript libraries to Python, added in Ver.7.7 in September 2026, is available only with the AI integration license

UiPath

  • Price revision: prices of all products rose by about 10% for contracts from July 1, 2023, including renewals
  • Licensing: on October 31, 2023, UiPath stopped offering all on-premises (legacy) plan licenses and Non-Production Robot licenses, so renewals must move to the Flex plan

Power Automate

  • Desktop version: running desktop flows on your own PC, started by a person and signed in with a Microsoft account, costs nothing extra
  • Organizational use: starting flows from the cloud, or running them with nobody present, needs an organizational Premium account (equivalent to ¥2,248 per user per month, billed annually, excluding tax) or other licenses

End of browser support

Support for the Internet Explorer 11 desktop application ended on June 15, 2022, and it was permanently disabled on February 14, 2023. Bots that operated IE had to be rebuilt for Edge or another browser. In a WinActor customer case, Persol Tempstaff planned an 11-month schedule to move all the scenarios for about 300 business processes to Edge.

Rework like this costs in proportion to the number of bots, and it arrives all at once every few years.

How to total it

First, collect these figures for each bot. Rough estimates from the owners’ memory are fine:

  • How many times it stopped in the last year
  • How long each stop took to investigate and fix
  • How long each stop took in manual work until it was fixed
  • Time spent updating the ledger and procedures, and on inventories
  • What you paid outside vendors for development and maintenance

Then total the annual running cost as follows:

Cost How to calculate
Licenses The annual contract amount
Environment The annual cost of PCs, servers, and accounts for bots
Fixing Stops × hours to fix each × hourly rate + outside maintenance fees
Downtime Stops × hours of manual work each × hourly rate
Management Hours on ledgers, procedures, and inventories × hourly rate
Rework and migration The cost of the last major rework ÷ years until the next

For the hourly rate, use the owner’s salary plus the employer’s share of social insurance and similar costs.

A worked example

Here is an example with made-up figures. Say there are 20 bots, each stops 4 times a year, each stop takes 3 hours to fix and 2 hours of manual work until fixed, and the hourly rate is ¥4,000.

  • Fixing: 20 bots × 4 stops × 3 hours × ¥4,000 = ¥960,000
  • Downtime: 20 bots × 4 stops × 2 hours × ¥4,000 = ¥640,000

Apart from licenses, stops alone take ¥1.6 million worth of time a year. Ledger and procedure updates, and the major rework every few years, come on top.

When you actually do the math, the cost is often concentrated in a handful of bots that stop often. Those few are the first candidates to replace.

What happens to costs when you replace bots with AI

Automation in which AI reads and operates the screen stops less often on small screen changes, which reduces the cost of fixing and downtime. Depending on how it works, though, other costs grow:

  • AI that reasons through every run: each run costs AI model fees and takes time. Gartner says that for processes run at high frequency, AI agents can cost up to 30 times as much as RPA bots
  • AI that learns an operation once and repeats it: AI is used only to learn an operation, relearn it after a screen change, or read information from the screen. Repeating the clicks and entries it learned costs no AI fees

When you compare, compare the totals above, not license fees.

The order to cut costs

  1. Retire unused bots: remove the licenses and environments of bots whose work is gone, found in your inventory
  2. Replace with official integrations: if a system offers an API or CSV download, stop operating its screens
  3. Replace the bots that stop most: move the few bots where cost is concentrated to automation that is easy to fix when screens change
  4. Keep track with a ledger: measure what you pay for, and how much, with the same yardstick every year

If a license renewal is near, plan the replacement backward from it. See How to switch from RPA to AI RPA for the steps, and How to build a robot ledger for RPA for the ledger.

How Kitewell is priced

Descarty’s Kitewell is priced by the computers it runs on, never by how many people use it. Plans for organizations, Team included, are quoted on the number of computers plus an annual base fee, for your scale and the management features you need.

Clicks and entries that worked are remembered and repeated, so AI is not called for every operation. AI is used to learn an operation, to relearn it after a screen change, and to read information from screens or documents, such as invoices. Besides cloud AI APIs, you can use in-house or local models, though small local models usually cannot operate a screen.

You can start by choosing one robot that keeps stopping and testing its replacement in a two-week, ¥300,000 (excluding tax) PoC. If the agreed success criteria are not met, you pay nothing. See Pricing & PoC for details.

Summary

  • RPA’s running cost is the sum of licenses, environment, building, fixing, downtime, management, and rework and migration
  • Fixing, downtime, and management are the hard-to-see costs, and they are almost entirely people’s time
  • The main RPA products keep revising prices and licensing, and VBScript and browser support keep ending
  • Collect each bot’s stops and hours, and you will find the few bots where cost is concentrated. Replace those first

Sources

← All articles

From automation that breaks to automation you can fix

Tell us which work you have in mind, and we will propose how a PoC would run.

Request a 15-minute demoDownload the overview