What outsourcing automation costs: hand-off vs guided

When you ask someone outside to automate work, the quotes you compare are for the first build. The real difference in cost shows up afterwards. A customer’s website changes, a form gains a column, the person in charge moves on. Who fixes the automation each time, and what it costs, decides the total for the year.

This column splits the ways of automating into three and compares the first cost, the monthly cost, and the fixes after launch. The market figures are for Japan.

Key points

  • Handing off one data-entry or transcription task typically costs about ¥50,000–150,000 in Japan. Each change to a screen or form then needs a maintenance contract or a fresh quote
  • Guided in-house automation means your own staff build it with outside help. Because fixes stay in-house, you stop paying outside for every change
  • Guidance does not suit a company that cannot name even one person to own the work

Three ways to automate

Hand it off Build in-house with guidance Buy a tool and build it yourselves
Who builds The contractor Your staff, with outside support Your staff
First cost About ¥50,000–150,000 per task; about ¥200,000–500,000 across several systems Support fee plus the tool The tool, plus the time your staff need to learn
Monthly cost About ¥20,000–100,000 for maintenance, plus the tool The tool The tool
When a screen or form changes The contractor fixes it (contract or fresh quote) Your staff fix it Your staff fix it
What you keep A working automation A working automation, and people who can fix it People who can fix it, once they learn
Suits Companies that cannot assign anyone, with work that won’t change soon Companies that can assign someone without automation experience Companies that already have someone who can build

The hand-off figures come from a 2026 Japanese article on outsourcing costs (see Sources) and vary widely with the work and volume.

The cost gap opens after launch

Automation is not finished when it is built. A customer’s website update, a replaced internal system, a changed form layout: each can stop it (why RPA robots stop when screens change).

When you hand it off, fixing is the contractor’s job: within a monthly maintenance contract, or quoted each time without one. A maintenance contract of ¥20,000–100,000 a month comes to ¥240,000–1,200,000 a year. Until a fix lands, the work goes back to hand.

When your own staff can fix it, the cost is their time. But if only the builder understands it, you end up worse off than with a contractor the day that person moves on (taking over RPA robots whose builder has left). Building in-house only works if what you build can be read by anyone.

For the full total, licenses and fixing effort included, see what RPA really costs to keep running. AI agents also need fixing when their models retire; Building AI agents in-house compares outsourcing, FDEs, and in-house by how fast each can fix them.

Tool costs in Japan (as of October 2026)

Tool Cost
Power Automate Premium about ¥2,248 per user per month (billed annually, excluding tax)
WinActor Node-locked full edition ¥1,098,680 a year; runtime edition ¥300,080 a year (per PC, including tax)
Kitewell For companies, agreed individually as an early access partner

What it takes to run a flow at a set time with the free version of Power Automate Desktop is covered in scheduling Power Automate Desktop.

How guided in-house automation runs

Here is how Descarty does it.

First, a joint verification

Pick robots that keep stopping, or manual tasks, and run them as Kitewell workflows as an early access partner. In the joint verification, Descarty builds the workflows with your team and measures the result.

  • It takes one month and costs ¥500,000 before tax, with no charge if the agreed success criteria are not met
  • At the end, part of the target screen is changed to see whether the workflow keeps working unchanged
  • The report covers items processed, items that succeeded, time per item, AI usage cost, and how often and why it stopped
  • AI model usage can be paid the way your requirements call for, including on Descarty’s invoice
  • The results are published as a joint verification case study, without your company name if you prefer

After the joint verification, your staff build

Your staff build the production work by talking with Kitewell’s assistant. Descarty’s build support helps choose what to automate, design the exceptions and approvals, and get unstuck along the way. Every workflow can be read as a diagram and as YAML, so people other than its builder can check it.

See Early access partners for the program details.

Who this does not suit

Guided in-house automation is the wrong fit when:

  • No one can own it: guidance exists to leave people in-house who can fix the automation. If the people who own the work cannot spend time on it, a hand-off with a maintenance contract fits better
  • Both systems have APIs: to connect cloud services to each other, an integration service such as Zapier is cheaper and steadier
  • You only want invoices received and read: a cloud service built for receiving invoices suits that better. Kitewell helps with what comes after: entering and matching them in internal systems that have no API
  • The site’s terms forbid automated use: don’t automate it (check the terms before automating)

What to ask when you get a quote

Whichever way you go, ask these to see the costs that come later:

  1. Are fixes for changed screens or forms included in maintenance? If not, what does each one cost?
  2. Can your staff read what was built? Can it be handed to another contractor?
  3. When it stops, who notices, and how?
  4. How does the price change as volume or the number of tasks grows?
  5. If it uses AI, who pays for the AI usage?

Summary

  • Outsourcing costs differ less in the first build than in the fixes after it
  • If you can assign one person, building with guidance and keeping people who can fix it saves paying outside for every change
  • If you cannot, choose a hand-off with a maintenance contract and agree in writing what fixes it covers

Sources

Verify it on your own work as an early access partner

We run Kitewell workflows with you on the same screens and data as production, and measure processing time, success rate, and how easily a stopped run is fixed.

The joint verification differs by engagement. For system integration, we set the requirements and design and test on production screens through errors and screen changes. For adoption by your staff, it is a one-department pilot: staff are trained, build their own tasks, and the effect is measured.

What we ask of you

  • One contact person for the work being automated
  • Procedure documents or manuals (or a recording of the work if there are none)
  • Test accounts and access to the screens involved
  • Sample data close to production
  • A machine to run Kitewell
  • An API key for the AI model you will use, or an in-house model server
  • Regular feedback during the verification
  • Help publishing the joint verification case study
3 companies, until March 31, 2027

Verify Kitewell with us on your own work

  • Kitewell's developers support you directly, from the first interview to production
  • Features your work needs are added to Kitewell itself, with priority
  • The results are published as a joint verification case study, without your company name if you prefer
  • One month for ¥500,000 before tax, with no charge if the agreed success criteria are not met
  • If you go on to a contract, the verification fee is credited to the first year

See the program details