RPA cost: what robotic process automation costs to implement and run
RPA cost, the cost of robotic process automation, has three parts: licences for the platform and the bots, the implementation work to map, build and test each process, and the maintenance that keeps bots working when the applications they click through change. The licence is the part on the price list; implementation and maintenance are usually the larger and less predictable parts. This guide explains each cost driver, compares RPA with API integration and custom automation over several years, lists the costs people forget and shows how to compare quotes.
It does not quote our own prices. It explains how an RPA budget is built, so any quote you receive, from us or from anyone else, makes sense. For how we approach automation, see business process automation; for what RPA is good at, see our guides to the benefits of RPA and RPA use cases.
The short version
- Licences: the platform, each bot, and orchestration or AI add-ons, every year.
- Implementation: process mapping, building, testing and documenting each bot.
- Infrastructure: machines or cloud capacity for unattended bots, credentials and monitoring.
- Maintenance: fixes when screens, fields or logins change, and exception handling.
How RPA licences are priced
RPA vendors such as UiPath, Microsoft Power Automate and Automation Anywhere price differently and change their plans often, so check their pricing pages on the day you compare. The common building blocks are:
| Licence type | What it covers | What to watch |
|---|---|---|
| Attended bot or user | A bot that runs on an employee's computer when they start it | Priced per user; grows with the team |
| Unattended bot | A bot that runs on its own on a server or virtual machine | Each one runs one task at a time; volume needs more bots |
| Orchestration | Scheduling, queues, credentials and monitoring for many bots | Often required once more than a few bots run |
| Document or AI add-ons | Reading invoices, forms and emails, or language models in a process | Often priced per page or per request |
| Usage-based plans | Pay per run or per minute of bot time | Check the cost at your busiest month |
Implementation is mostly people's time
Every bot is built from hours: watching the process as it is really done, writing down every exception, building, testing with real cases and documenting what the bot does. What an hour costs follows where the team works. Eurostat's release of 31 March 2026 puts average hourly labour costs across the whole economy in 2025 at €34.9 in the EU and €38.2 in the euro area, from €12.0 in Bulgaria to €56.8 in Luxembourg (read on 12 November 2026). Compare totals for the same processes, not hourly rates. The time your own staff spend explaining the process and checking results is a real cost too.
What drives RPA cost per process
| Driver | Why it adds work | How to keep it in check |
|---|---|---|
| Steps and applications | Each application the bot opens adds logins, screens and failure points | Automate the part of the process inside one or two systems first |
| Exceptions | Every case that does not follow the rule needs a branch or a hand-off to a person | Let the bot hand exceptions to staff instead of solving them all |
| Input format | Structured data is simple; scanned documents and free-text emails need reading and checks | Standardise the input before automating it |
| Screen stability | Web pages and applications that change often break bots | Prefer stable systems, or an interface where one exists |
| Volume and timing | High volumes or deadlines need several unattended bots and queues | Measure volume per day and peak before sizing |
| Controls | Finance and personal data processes need logs, approvals and access rules | Agree the controls with the process owner and auditor early |
RPA, API integration or custom automation
RPA is often the right tool when the systems have no usable interface, when the process is stable and rule-based, and when results are needed quickly. A good supplier will also say when it is not. Bots that work through screens break when the screens change, so where a system offers a stable interface, an API integration or a custom automation service usually costs less to keep running. Compare three to five years for the same processes: licences, infrastructure and bot maintenance against the build and upkeep of an integration. Many companies use both: RPA for older systems, interfaces for the rest.
When RPA meets AI: the AI Act
Rule-based bots are generally not AI systems, but RPA is increasingly combined with document reading and language models. When it is, the AI Act can apply to that part. Regulation (EU) 2024/1689 applies generally from 2 August 2026 (Article 113). After Regulation (EU) 2026/1744, the high-risk rules apply from 2 December 2027 for systems listed in Annex III, which includes AI used to filter job applications or to evaluate creditworthiness, and from 2 August 2028 for AI in products covered by Annex I. Article 4, as amended by Regulation (EU) 2026/1744, requires providers and deployers to take measures to support the development of AI literacy of their staff. Separately, the GDPR gives people the right not to be subject to a decision based solely on automated processing that significantly affects them, with exceptions (Regulation (EU) 2016/679, Article 22; all read on 12 November 2026). These rules add design and documentation work; they belong in the quote.
The hidden costs of RPA
- Bot maintenance. A new field, a changed login page or an application update can stop a bot overnight.
- Licence growth. Each new process can mean another bot or a higher plan.
- Exception handling. Cases the bot cannot handle still need staff, and a queue to put them in.
- Credentials and access. Bots need their own accounts, managed and reviewed like any user.
- Infrastructure. Machines for unattended bots, and someone to keep them patched.
- Governance. Twenty bots without owners and documentation become a risk of their own.
How to compare RPA quotes
- Which processes are included, by name, with volumes and exceptions?
- Which licences are assumed, and what do they cost when you add processes?
- Who provides and maintains the machines for unattended bots?
- How are exceptions handled and reported?
- What does maintenance cover when screens change, and how fast?
- Is each bot documented so another team can maintain it?
- Where a system has an interface, has the supplier compared RPA with integration?
- Are the bot code, credentials and logs in your company's accounts?
Ways to spend less without cutting quality
- Fix the process first. Automating a poor process makes the poor process faster.
- Start with stable, high-volume processes. They pay back first and break least.
- Hand exceptions to people. A bot that does 80% reliably beats one that tries everything.
- Use interfaces where they exist. Fewer screens mean fewer breakages.
- Own the automation. Documented bots in your accounts keep every future change open.
Related reading: process mapping, software maintenance cost, AI agent development cost and RPA in banking. All our services are on the services page, and every guide is in the guides library.
Questions people ask
How much does RPA cost?
RPA cost has three parts: licences for the platform and each bot, the implementation work to map, build and test each process, and the maintenance that keeps bots running when the applications they work in change. Licences are priced by the vendor, often per bot or per user; implementation and maintenance follow the number and complexity of processes. Any figure given before the processes are mapped is a guess.
How much does RPA implementation cost per process?
It depends on the number of steps and applications, how many exceptions the process has, how stable the screens are and how much testing and documentation the process needs. A simple, rule-based process with structured input is a small project; one with many exceptions, unstructured documents and several systems is several times the work.
Is RPA cheaper than an API integration?
At the start, often yes, because a bot can work through existing screens without changes to the systems. Over time an API integration usually costs less to keep running, because interfaces change less often than screens. Where a stable interface exists, compare both over three to five years.
What are the hidden costs of RPA?
Bot maintenance when screens, fields or logins change, licence growth as more processes are automated, exception handling by staff, infrastructure for unattended bots, credentials management, monitoring and the governance of a growing number of bots.
Does the AI Act apply to RPA?
Rule-based bots that click through screens are generally not AI systems. When RPA is combined with AI, such as document reading or a language model making decisions, the AI Act can apply to that part. It applies generally from 2 August 2026, and the high-risk rules for Annex III uses such as recruitment or credit scoring apply from 2 December 2027 after Regulation (EU) 2026/1744.
Sources
- Eurostat: EU hourly labour costs ranged from €12 to €57 in 2025 (31 March 2026, read 12 November 2026)
- Regulation (EU) 2024/1689 (AI Act), Article 113 and Annex III (read 12 November 2026)
- Regulation (EU) 2026/1744 amending the AI Act (Article 4, dates for high-risk rules) (OJ 24 July 2026, read 12 November 2026)
- Regulation (EU) 2016/679 (GDPR), Article 22 (read 12 November 2026)
Your RPA scope checklist
Write one line for each before you request quotes. The answers make quotes comparable.
- The processes to automate, each in one sentence.
- Volume per day and at the peak, for each process.
- The applications each process touches, and whether they have an interface.
- The input: structured data, documents or emails.
- Known exceptions, and who handles them today.
- Attended or unattended bots.
- Controls: approvals, logs and access rules.
- Any AI in the process, and what it decides.
- Who owns each bot after launch.
- Ownership: bot code, credentials and logs in your company's accounts.