Skip to content
Published on

RPA Platforms 2026 Complete Guide - UiPath, Automation Anywhere, Blue Prism, Power Automate, Tungsten, ABBYY, Rocketbot Deep Dive

🇰🇷KO🇺🇸EN🇯🇵JA
Split View필사 모드
Authors

Prologue — Is RPA Dead, or Reborn, in 2026?

Around 2018, when the word "RPA" first showed up in enterprise meeting rooms worldwide, it was a magic word. Excel work, ERP screen clicking, copying data out of email attachments — bots would do all of it. UiPath IPO'd in April 2021 at a $35 billion market cap.

Fast forward to 2026, and the landscape has flipped. Forrester's 2024 report warned of a "slowdown in traditional RPA growth", and Gartner's 2025 Hype Cycle pushed RPA back near the Trough of Disillusionment rather than the Slope of Enlightenment. And yet — paradoxically — UiPath crossed $1.3 billion in revenue in FY2025, Anthropic shipped Computer Use that same year, and OpenAI GA'd Operator.

RPA in 2026 is no longer "screen scraping + rule-based workflow." It is an "automation stack where LLMs decide, vision models see, and agents act." UiPath Autopilot, Automation Anywhere Co-Pilot, Power Automate Copilot — every major vendor is running in the same direction.

What this article covers:

  1. What RPA is and where it stands in 2026
  2. UiPath 25.4 — Autopilot, Orchestrator, AI Center, Document Understanding
  3. Automation Anywhere 360 — AARI, Document Automation, Co-Pilot
  4. Blue Prism (post-SS&C acquisition) — Decipher IDP, Cloud Foundation
  5. Microsoft Power Automate — Copilot, AI Builder, Desktop Flows
  6. Tungsten Automation (formerly Kofax) — TotalAgility, RPA, AP Essentials
  7. ABBYY — FineReader Server, Vantage, IDP focus
  8. Rocketbot — Latin America's rising star
  9. WorkFusion, Pega, NICE, OpenRPA, TagUI
  10. Python automation alternatives — pywinauto, PyAutoGUI, robotframework
  11. AI-first agents — Anthropic Computer Use, OpenAI Operator, Vertex AI Agents
  12. Browser automation rise — Playwright, Browser Use, Skyvern, Multion
  13. Document understanding (IDP) — Textract, Form Recognizer, Document AI, Hyperscience
  14. Korean RPA — Samsung SDS Brity RPA, AhnLab DRA, LG CNS, Naver Clova, ZAPER
  15. Japanese RPA — WinActor, BizRobo!, BlueX, SoftBank
  16. Pricing models — per bot, per workflow, attended vs unattended
  17. The top ten RPA processes
  18. The RPA decline narrative — Forrester downgrade, Gartner Hype Cycle
  19. The future — Agentic AI, Hybrid RPA+LLM, computer vision
  20. Which platform to choose
  21. References

1. What RPA Is, and Where It Stands in 2026

The original definition of RPA (Robotic Process Automation) is simple: "software bots that mimic the repetitive GUI work humans used to do." Open the ERP, fill in fields, click Save, jump to the next row — done by a bot.

In the early 2010s, Blue Prism created the category in the UK BPO market. UiPath, founded in Bucharest, Romania in 2005, started exploding in 2017. Automation Anywhere grew in San Jose during the same era. These three came to be known as the "RPA Big 3."

But in 2026, the definition itself is shaky. Three reasons:

One — the fundamental fragility of UI automation. Shift a button by one pixel and the bot breaks. ERP updates, browser version bumps, OS patches — the bot needs to be rewritten each time. Maintenance cost often exceeds adoption cost.

Two — the LLM arrival. GPT-4 and Claude 3 made "unstructured data" tractable. Pulling an invoice number from email body, reading a table out of a PDF — what used to require regex + OCR pipelines is now a single LLM call.

Three — AI agents. Anthropic's Computer Use (October 2024), OpenAI's Operator (January 2025), Google's Vertex AI Agent — these tackle RPA territory by having an "LLM look at the screen and click directly." No rules to write. Just give it a goal.

These three forces have combined to push every major RPA vendor in 2026 to rebrand: "We are no longer an RPA company, we are an Intelligent Automation / Agentic Process Automation company." UiPath changed its tagline from "Automation" to "AI at Work." Tungsten Automation literally renamed itself from Kofax.

One-line state of RPA in 2026: Pure rule-based RPA market is flat, AI-fused Intelligent Automation is double-digit growth, and above both there is a bigger category — AI Agents — quietly absorbing RPA.

2. UiPath 25.4 — The Reigning Champion's Evolution

UiPath is still the RPA market share leader. FY2025 (ending Jan 2025) revenue: 1.34billion.ARR:1.34 billion. ARR: 1.67 billion. Leader spot in Forrester Wave Q1 2024. Top-right in Gartner Magic Quadrant 2024.

UiPath Platform 2024.10 / 25.4 — what changed

The UiPath product suite consolidated from 2024 onward looks like this:

  1. Studio / Studio Web — Developer IDE. Studio is Windows desktop; Studio Web runs in the browser. Studio Web has been eating the citizen-developer market since launch in 2023.
  2. Orchestrator — Bot management, scheduling, monitoring central server. Available as cloud (Automation Cloud), on-prem, or hybrid.
  3. Robots — The actual automation runtime. Attended (next to user), Unattended (headless server), Hybrid.
  4. AI Center — ML model hosting and management. Both custom-trained and pre-trained models.
  5. Document Understanding — IDP (Intelligent Document Processing). Invoices, contracts, IDs, receipts to structured data.
  6. Communications Mining — Pulls intent out of email, chat, tickets. Based on the Re:infer acquisition from 2022.
  7. Process Mining — Analyzes system logs to discover processes worth automating. Based on Cloud Elements and StepShot acquisitions in 2022.
  8. Action Center — Human-in-the-loop. A task queue where bots escalate to humans.
  9. Apps — Low-code app builder. Builds UIs that trigger bots.
  10. Insights — BI and dashboards.
  11. Test Suite — Bot test automation.

The big 2025 shift — Autopilot

UiPath Autopilot, announced October 2024, is the GenAI copilot. It comes in three forms.

  • Autopilot for Studio — Generates workflows from natural language. Type "build me a bot that pulls amounts from invoice PDFs and posts them to SAP" and it drafts the workflow.
  • Autopilot for Testers — Auto-generates test cases.
  • Autopilot for Everyone — End users invoke bots via chat. Combined with Action Center.

UiPath's big bet is Agentic Automation. In a November 2024 blog post, the company declared "a future where agents and bots work together," and added UiPath Agent Builder in 2025. Unlike code-first frameworks like LangGraph, this is a GUI-first way to define LLM agents that call RPA bots as tools.

Pricing (per 2026 public information)

UiPath is fundamentally per-robot. An Unattended Robot is typically 8,000–8,000–12,000 per year. Studio Pro starts at $1,930 per developer per year (UiPath public pricing, negotiable). Real enterprise contracts are usually rolled into ELAs (Enterprise License Agreements).

Field note: A common refrain from Korean enterprise adopters is "the first three years are fine, but year four renewal negotiations get hard." Meanwhile, Document Understanding and Communications Mining remain hard for competitors to match.

3. Automation Anywhere 360 — Cloud-Native Pride

Automation Anywhere once stood shoulder-to-shoulder with UiPath after a 550M SoftBank Vision Fund check in 2018. In 2026 it sits at #2 or #3 market share, still private, with estimated revenue around 700 million.

Automation 360 (previously Enterprise A2019) — 2026 lineup

  1. Automation Anywhere Control Room — Bot orchestration. UiPath Orchestrator analog.
  2. AARI (Automation Anywhere Robotic Interface) — User interface for attended automation, introduced in 2020.
  3. IQ Bot → Document Automation — IDP. Renamed in 2024.
  4. Automation Co-Pilot — GenAI assistant. UiPath Autopilot analog.
  5. AI Tools (formerly AI Hub) — Gateway integrating LLMs, NLP, OCR.
  6. Process Discovery — Process Mining product.
  7. CoE Manager — Governance tooling for Center of Excellence teams.

Technical character

Automation Anywhere's biggest differentiator is its Web-First architecture. Bot development itself happens in the browser (UiPath's Studio Web has chased the same direction). Every feature is cloud-native, multi-tenant by design from day one.

The peculiarity of Document Automation

Automation Anywhere's Document Automation supports self-trained ML models. Like UiPath Document Understanding, label 5–10 invoices and the model learns. Starting in 2025, the company added its own Vision-Language Model (IQ Document AI).

What sets Automation Co-Pilot apart

If UiPath Autopilot is an IDE-integrated copilot, Automation Anywhere's Co-Pilot is squarely aimed at end-user business users. It sits as a sidebar next to Salesforce, SAP, ServiceNow and similar SaaS UIs, and triggers bots from natural language.

Pricing

Similar to UiPath. Unattended Bot 7,500–7,500–11,000 per year. Plus AARI user licenses on top.

Market view: Automation Anywhere shelved its IPO attempt in 2024 and stayed private. Its more aggressive cloud-first strategy and slicker SaaS UX are real strengths, but share in Korea and Japan trails UiPath.

4. Blue Prism — Identity After SS&C

Blue Prism is the UK company credited with inventing the RPA category. Founded in 2001, listed on the LSE in 2016, then acquired by SS&C Technologies in April 2022 for $1.24 billion. Now branded SS&C Blue Prism.

Blue Prism's differentiator — Server-First Architecture

While UiPath and AA started with desktop automation, Blue Prism was built from day one around headless server bots. Attended bots were added much later. That made the platform a favorite in financial services and BPO — where security and governance trump everything.

2026 product suite

  1. Blue Prism Enterprise / Cloud — Core RPA platform. Process Studio + Object Studio + Control Room.
  2. Decipher IDP — IDP solution, announced 2021.
  3. Blue Prism Cloud Foundation — Azure-based managed cloud option.
  4. Capture & Process Intelligence — Process Mining offering.
  5. Chorus BPM (SS&C synergy) — Workflow + RPA combined.

What changed after SS&C

Upside: combined with SS&C's deep financial-BPO customer base. Downside: the market perceives slower innovation as an independent vendor. Through 2024–2025, the GenAI copilot announcements lagged UiPath, AA, and Microsoft.

Technical character

Blue Prism still has you build bots in Process Studio with BPMN-like diagrams. One page for business logic, another for UI objects. That separation remains the most-praised maintainability advantage.

Pricing

Unattended Bot 10,000–10,000–15,000 per year — the priciest of the bunch. The flip side is a famously short free trial in PoC stages.

Field note: Blue Prism deployments are less common in Korea compared to UiPath/AA. In Japan, it retains some footprint, mostly in financial services.

5. Microsoft Power Automate — The Dark Horse Counterattack

Microsoft Power Automate started life as Microsoft Flow, a lightweight integration tool. Power Automate Desktop arrived in 2019, marking the proper entry into RPA. By 2026 it is positioned to threaten the #1 slot.

Why Power Automate is growing scarily fast

  1. M365 bundling — Power Automate is included in Microsoft 365 licensing. You can do desktop automation without buying a separate RPA license.
  2. Azure integration — Connects naturally to Azure OpenAI, Azure AI Document Intelligence, Logic Apps.
  3. Copilot — Microsoft Copilot's RPA extension. Generate flows from natural language.
  4. Cheap entry pricing — Where UiPath/AA charge 10Kperbot,PowerAutomatestartsat10K per bot, Power Automate starts at 15 per user per month.

2026 product suite

  1. Cloud Flows — SaaS-to-SaaS integration (Zapier analog). 1000+ connectors.
  2. Desktop Flows — Desktop RPA (UiPath/AA analog).
  3. Process Mining — Built on the Microsoft Minit acquisition.
  4. AI Builder — Low-code ML model builder for form processing, classification, prediction.
  5. Copilot for Power Automate — Natural-language flow creation. "Save email attachment PDFs to SharePoint every Monday at 9am."

Power Automate Desktop's technical ceiling

Power Automate Desktop matches UiPath/AA at the basic Windows desktop automation level. But enterprise governance and large-scale bot operations still trail. Sophisticated Orchestrator-class bot queue management is weaker.

Pricing (2026)

  • Power Automate Premium — $15 per user per month. Desktop bot usage included.
  • Power Automate Process — $150 per bot per month (Unattended).
  • Power Automate Hosted Process — $215 per bot per month (Microsoft-hosted).
  • AI Builder Credits — Separately metered. Starts at $100 per month.

A common assessment: roughly one-fifth the UiPath sticker price.

Market view: Forrester upgraded Microsoft from "Strong Performer" to "Leader" in its 2024 RPA report. Estimates suggest Power Automate captures up to 30% of new RPA deployments by 2026.

6. Tungsten Automation (formerly Kofax) — IDP Old Guard

Kofax was founded in 1985 — the original heavyweight in document capture. Acquired by Thoma Bravo private equity in 2022, then renamed Tungsten Automation in January 2024. Around the same time, some business units were rolled together with Tungsten Network.

Tungsten Automation 2026 portfolio

  1. TotalAgility — Core integration platform. Intelligent Document Processing + RPA + BPM, unified.
  2. Tungsten RPA (formerly Kofax RPA) — UI automation bots.
  3. Tungsten Capture — Paper, image, PDF capture.
  4. AP Essentials — Accounts Payable SaaS. Specializes in invoice processing.
  5. SignDoc — E-signature.

Why Tungsten survives

If UiPath/AA started "RPA-first," Tungsten started "capture-first." In 2026, more than 60% of all RPA projects involve documents — invoices, IDs, contracts. In that area, Tungsten has 30 years of know-how.

Tungsten Vantage

Like ABBYY's, an IDP-focused product. Next-generation IDP with Vision Language Model integration. Announced 2024.

Pricing

Typically hybrid licensing: per page or per document for capture, per bot for RPA. AP Essentials is priced per invoice.

Field note: For AP (accounts payable) automation specifically, Tungsten AP Essentials competes with SAP Concur and Tipalti. Better viewed as IDP market more than general RPA.

7. ABBYY — From FineReader Server to Vantage

ABBYY was founded in Moscow in 1989 — an OCR/IDP powerhouse. (Headquarters moved to the US after 2022.) Consumers know FineReader PDF; enterprise has a separate lineup.

ABBYY 2026 portfolio

  1. FineReader Server — High-volume OCR processing server. PDFs and images turn into searchable PDFs or text.
  2. Vantage — Next-generation IDP platform. Launched 2020. Pre-trained "Skill" marketplace (invoices, contracts, IDs) speeds adoption.
  3. Timeline (formerly Process Intelligence) — Process Mining product. Acquired 2019.
  4. FlexiCapture — Legacy IDP platform. Vantage is the successor.

ABBYY's value proposition

ABBYY is a pure IDP vendor. It does not sell RPA bots directly. Instead, it ships integrations to UiPath, Automation Anywhere, and Power Automate. The strategy: "pick your RPA vendor, but use ABBYY for document recognition."

Pricing

Vantage is typically priced per page. Approximately 5–25 cents per invoice page (negotiable).

Market view: As LLMs and VLMs evolved in 2025, IDP vendor differentiation has narrowed. ABBYY's "dozens of pre-trained Skills" library, however, remains a real asset.

8. Rocketbot — Latin America's Dark Horse

Rocketbot was founded in Chile in 2017. Largely unknown in Korea and Japan, but rapidly capturing share in Latin America (Brazil, Mexico, Colombia, Argentina). Recognized as a Strong Performer in the 2024 Forrester Wave.

Rocketbot's differentiators

  1. Low-cost positioning — 50–70% cheaper than UiPath/AA.
  2. Spanish/Portuguese-first — UX designed for the Hispanic market from the start.
  3. Rocketbot Studio — Python-based scripting. Developer-friendly.
  4. Official marketplace — Plenty of pre-built bot templates.

Technical character

Rocketbot lets you export bots as Python scripts — the biggest divergence from other major RPA. It is closer to "RPA where you write code."

Market view: Limited footprint in Korea and Japan, but undeniable in price-sensitive markets (SMBs, Southeast Asia, Latin America). 2026 revenue estimated around $50 million.

9. WorkFusion, Pega, NICE, OpenRPA, TagUI

This is the category beyond the RPA Big 3 + Microsoft + Tungsten + ABBYY.

WorkFusion — Founded 2010, New York. AML (anti-money laundering) RPA specialist. Sells pre-trained bots branded as "Digital Workers." Strong in financial KYC/AML.

Pega — Pegasystems' RPA module. Embedded in the Pega Infinity BPM/CRM platform. Not really a standalone RPA vendor — more "existing Pega customers also use RPA."

NICE Robotic Automation — RPA from Israel-based NICE. Specialized in contact-center automation (customer service). NEVA (Nice Employee Virtual Attendant) is the attended-bot brand.

OpenRPA — Open-source RPA originating in Denmark. MIT license. Workflow diagram UI similar to UiPath. Self-hostable. Small but active community.

TagUI — Open-source RPA built by AI Singapore. Apache 2.0. Bots are written using natural-language commands (English sentences) — an unusual choice. An early swing at RPA + LLM convergence.

Robocorp — Helsinki, Finland. Open-source RPA built on Robot Framework. Python-first. Rebranded to Sema4.ai in 2024 as it pivoted into AI agents.

10. Python Automation Alternatives — pywinauto, PyAutoGUI, robotframework

If you don't want to buy RPA vendor licenses, or you have a developer team that prefers code, the Python ecosystem has serious alternatives.

pywinauto — Windows desktop automation. Supports both UIA (UI Automation) APIs and Win32 messages. The control-tree traversal is conceptually similar to UiPath's Object Repository.

from pywinauto import Application
app = Application(backend="uia").start("notepad.exe")
app.UntitledNotepad.Edit.type_keys("Hello", with_spaces=True)
app.UntitledNotepad.menu_select("File -> Save As")

PyAutoGUI — Cross-platform GUI automation. Pixel-coordinate-based mouse and keyboard control, with image matching (template matching) to locate screen elements. Simplest of all, also the most fragile.

robotframework — Open-source test automation framework. Pair it with rpaframework and you get a full RPA platform. The technology underpinning Robocorp/Sema4.ai. Bots are written in keyword-driven English-like sentences.

Playwright — Originally a browser test tool, but in 2026 the fastest-growing tool in the RPA space. The de facto standard for headless browser automation.

Developer view: For simple tasks, Python automation is far cheaper and more flexible than RPA vendors. The hard part is maintenance by non-developers. "Who maintains this?" is the actual crux of the RPA-vs-Python decision.

11. AI-First Agents — Computer Use, Operator, Vertex Agents

A new category emerged in 2024–2025 that genuinely threatens RPA's territory. The label: AI-First Agent or Agentic Automation.

Anthropic Computer Use (October 2024) — An API addition to Claude 3.5 Sonnet that lets it look at the screen and operate mouse and keyboard. The user instructs in natural language: "Take the sum in this Excel file and enter it into SAP." No workflow defined. Claude sees the screen, decides, acts.

OpenAI Operator (January 2025) — A browser-automation agent launched as part of ChatGPT. Navigates websites on the user's behalf, fills forms, even completes purchases.

Google Vertex AI Agent Builder — Google Cloud's agent builder service, powered by Gemini. CRM integrations, database queries, function calling are first-class.

Microsoft Copilot Studio + Autonomous Agents — Microsoft announced its "Autonomous Agents" concept in November 2024. LLM agents that pair with Power Automate.

Why these threaten RPA

Traditional RPA: "The button's coordinates are (532, 891). Click it, then find the 'Next' button on the next screen." AI-First agent: "Enter this invoice PDF into SAP. Look at the screen and figure it out."

No rules to write. Bots don't break when the UI shifts. In theory.

Reality check

As of May 2026, AI-first agents are slow, expensive, and probabilistic.

  • 30 seconds to several minutes per task (RPA does it in 5–10 seconds).
  • 0.10–0.10–1 per task in LLM API costs (RPA is effectively free).
  • 5–20% task-failure rate on identical inputs (a well-built RPA bot is under 1%).

So 2026 reality is hybrid: RPA bots handle baseline automation; AI agents take exception handling, new tasks, and UI areas that change often.

Perspective: UiPath and Automation Anywhere have both released their own AI Agent products. They treat agents not as "the enemy here to kill RPA" but as "RPA's next chapter."

12. The Rise of Browser Automation — Playwright, Browser Use, Skyvern, Multion

About 70% of real RPA work happens inside a browser. SaaS, web ERPs, intranets — far more web than native desktop. The tooling here is exploding.

Playwright — Microsoft's browser automation library. Originally for E2E testing, also powerful for RPA. Supports Chromium, Firefox, WebKit. Bindings for TypeScript, Python, .NET, Java.

Browser Use — Open-source, released November 2024. Combines Playwright with LLM reasoning. Executes natural-language commands like "book the cheapest hotel for me from this site." 50K+ GitHub stars, growing fast.

Skyvern — YC W24. Combines LLM + computer vision + Playwright. Dual-model: SaaS and open source, focused on browser automation.

Multion — OpenAI Fund-backed in 2024. Browser-automation agent SaaS.

Anthropic Claude in Chrome — Announced fall 2025. A Chrome extension where Claude operates directly inside the browser. Could eventually subsume Computer Use.

2026 browser-automation landscape: Simple form filling — Playwright. Complex decision-making — Browser Use or Skyvern. Enterprise RPA integration — UiPath/AA web automation modules.

13. Document Understanding (IDP) — Textract, Form Recognizer, Document AI

A huge slice of RPA usage is document understanding. Turning PDFs, images, scans into structured data. The 2026 landscape here is rich.

AWS Textract — AWS OCR + IDP service. About 1.5 cents per page basic, 5 cents per page with analysis. Strongest automatic table detection.

Azure AI Document Intelligence (formerly Form Recognizer) — Microsoft Azure. Pre-trained models for invoices, receipts, IDs, contracts. Around 10 cents per page.

Google Document AI — Google Cloud. 1000+ pre-trained processors. Now integrated with Gemini for vision-language reasoning.

Hyperscience — US IDP specialist. Founded 2014. Y Combinator alum. Strong on handwriting.

Rossum — Czech IDP startup. Invoice processing focused SaaS. Strong in Europe.

Instabase — US. LLM-based unstructured-data processing.

Nanonets — India IDP SaaS. Aggressive pricing.

Sensible (formerly Sensible AI) — Developer-focused IDP API. SDK-first.

Direct LLM usage — Since 2025, using GPT-4o, Claude 3.5 Sonnet, or Gemini 2.0 vision capabilities to do IDP without a dedicated vendor has exploded. "Why pay for IDP SaaS when I can throw the PDF at an LLM?" became a common question.

2026 IDP theme: Value of pre-trained skills vs. cost-effectiveness of general LLMs. Standardized invoices and receipts: LLMs are enough. Complex medical records and legal contracts: dedicated IDP still wins.

14. Korean RPA — Brity, DRA, Auto Roman, Clova, ZAPER

In Korea, global vendors (UiPath, Automation Anywhere) hold the top spots, but domestic vendors have a real presence too.

Samsung SDS Brity RPA — Samsung SDS's RPA platform. Widely adopted across Samsung group companies. Estimated #1 or #2 in domestic RPA share. Brity Automation is the umbrella brand combining RPA, Chat, and Workflow. GenAI integration added in 2024.

AhnLab DRA (Digital Robotic Automation) — RPA from AhnLab, the cybersecurity company. Security-friendly design is the selling point. Strong in financial and public sectors.

LG CNS Auto Roman — LG CNS's RPA, centered on LG group companies. Bundled with proprietary IDP.

Naver Clova OCR — Naver Cloud's OCR API. Top-tier Korean OCR quality. Per-page pricing. The most common IDP backend behind Korean RPA projects.

Kakao Enterprise OCR — Kakao Enterprise. Korean OCR competitor.

MetaScale CONNECTO — Korean RPA vendor. Mid-market.

ZAPER — A Korea-native Zapier alternative. SaaS-integration and automation SaaS. Specialized in internal approval and workflow automation.

Others — Daewoong Pharmaceutical's in-house RPA, KB Kookmin Card's RPA automation cases, and similar are conference regulars.

On-the-ground Korea: Large enterprises typically run UiPath alongside Brity RPA. Mid-market gravitates toward Brity or Power Automate. Financial services tend to favor UiPath.

15. Japanese RPA — The Standalone Ecosystem WinActor Built

Japan is uniquely positioned in the global RPA market. It is essentially the only major market where the global Big 3 is not #1 — that slot belongs to NTT's WinActor.

WinActor — Built by NTT Advanced Technology and sold by NTT Data. Launched 2010. Roughly 40% share of the Japanese RPA market (per 2024 ITR research). The strengths: Japanese-language UI, specialization in Japanese enterprise workflows, and a deep Japanese SI partner network.

BizRobo! (formerly Kofax Kapow) — Sold in Japan by RPA Technologies. Based on Kofax tech. Approximately #2 share in Japan.

WinActor vs. global RPA

WinActor differs from global RPA in:

  • A Japanese UX designed from the ground up (distinct from UiPath's Japanese localization).
  • Specialized activities for Japanese approval workflows and hanko-seal automation.
  • Tight integration with Japanese ERPs (OBIC, Works, Galileopt).

BlueX — Japan-grown RPA. SMB market.

Robotic Process Automation by SoftBank — SoftBank's RPA service. Combines Automation Anywhere Japan distribution with in-house SI.

Autoブラウザ名人 (Auto Browser Meijin) — User Local's browser automation tool. Specialized for the Japanese market.

Japanese IDP — AnyOne, ABBYY Japan, NTT Data Wadoki, AI inside (DX Suite) all compete here.

On-the-ground Japan: WinActor's dominance is so strong that global vendors struggle to break in. UiPath Japan focuses on large enterprises and foreign-affiliated companies. Power Automate is rising fast via M365 bundling.

16. Pricing Models — Per Bot, Per Workflow, Attended vs. Unattended

RPA licensing varies by vendor. The main models:

ModelDescriptionRepresentative vendors
Per RobotAnnual license per botUiPath, AA, Blue Prism
Per UserMonthly per userPower Automate
Per WorkflowPer flowZapier, Power Automate Cloud Flows
Per Page / Per DocumentPer page for IDPTextract, Form Recognizer, Tungsten AP
Per Run / Per ExecutionPer bot executionSome Power Automate SKUs
Consumption-BasedUsage-based (CPU, memory)Some cloud RPA

Attended vs. Unattended

  • Attended Robot — Runs next to the user's desktop. User triggers. Cheaper (annual 2K–2K–5K).
  • Unattended Robot — Runs headless on a server 24/7. Pricier (annual 8K–8K–15K).
  • Hybrid — Switches modes. Increasingly the default.

True cost of adoption

License fees are typically only 20–30% of TCO. The remainder:

  • Implementation cost — SI consulting. Per bot 5K–5K–50K.
  • Maintenance — 30–50% of annual license. UI changes, exception handling.
  • CoE operations — Center of Excellence headcount. Large enterprises run 5–20 full-timers.

Real-world ROI: The "one bot replaces ten people" pitch is almost always wrong. Reality is closer to "a bot covers 60% of routine repetition, freeing up 30% of one person's time." Even so, ROI usually still pencils out.

17. The Ten Most Common RPA Processes

The use cases most frequently shown in vendor conferences.

  1. AP Automation (invoice processing) — Invoice PDF → OCR → ERP entry → approval. The #1 RPA use case. Tungsten AP Essentials and SAP Concur are SaaS alternatives.
  2. Customer Onboarding / KYC — Identity verification, account opening, credit checks. Core for banks and brokerages.
  3. Order-to-Cash — Order entry, invoice issuance, payment matching.
  4. HR Onboarding — Cross-system data entry for new hires (IT, HR, finance, benefits).
  5. Payroll — Payroll data aggregation, tax calculations.
  6. IT Service Desk — Password resets, user account creation, AD group additions.
  7. Compliance Reporting — Regulatory report generation (AML, GDPR).
  8. Insurance Claims — Insurance claim processing: document collection, validation, system entry.
  9. Healthcare Revenue Cycle — Insurance claims, EHR data entry.
  10. Supply Chain — Purchase order entry, inventory checks, shipment tracking.

Korea-specific use cases

  • Tax invoice issuance and procurement processing (NTS Hometax integration).
  • Four-major insurance filings.
  • Electronic approval (groupware) automation.
  • University admissions processing (document collection, system entry).

Sobering point: Many of the above tasks now have SaaS solutions that beat RPA in 2026. RPA is often a "second-best when you can't replace the underlying system."

18. The RPA Decline Narrative — Forrester Downgrade, Gartner Hype Cycle

Analyst tone shifted starting in 2024.

Forrester — "The Forrester Wave: Robotic Process Automation Q1 2024" explicitly cited slowing growth. The category itself was redefined as "Pure-play RPA stagnant, Intelligent Automation growing."

Gartner — In the 2024 Hype Cycle for Artificial Intelligence, "Generative AI" sat at the peak, "RPA" moved to Plateau of Productivity. In 2025, however, "Autonomous AI Agents" became the new peak, raising fresh questions about RPA's future.

IDC — 2024 Worldwide Intelligent Process Automation Software Forecast: RPA grows single digits, AI-powered Process Automation grows double digits.

Market performance

  • UiPath's stock dropped over 80% from IPO peak (2023 trough). Recovered in 2024–2025 but still below IPO price.
  • Automation Anywhere delayed its IPO twice.
  • Blue Prism went private after the SS&C acquisition.

Root causes

  1. Maintenance burden — Bot breakage frequency far exceeded early estimates.
  2. Governance problems — Citizen-developer-generated bot proliferation ("Bot Sprawl") got out of hand.
  3. LLM arrival — Same problems solved more flexibly.
  4. SaaS evolution — Integrated APIs eliminated entire RPA use cases.

Counterpoint — Saying "RPA is dying" is overblown. Legacy systems aren't going anywhere, so RPA isn't going anywhere. But "pure RPA vendors" are struggling, and "AI + RPA combined vendors" are the survivors.

19. The Future — Agentic AI, Hybrid, Computer Vision

Some predictions for RPA (or its successor) in 2026–2028.

1. The rise of Agentic Process Automation

UiPath Agent Builder, Automation Anywhere AI Agent Studio, Power Automate Autonomous Agents — every major vendor is heading toward a hybrid architecture where "LLM agents make decisions and RPA bots handle execution."

2. Standardization of Computer Use models

Anthropic Computer Use opened the category in 2024. OpenAI Operator and Google followed. By 2026, open-source models (a future Llama, say) are likely to provide the same capability. Standardization erases the comparative advantage of an RPA vendor's "UI automation engine."

3. Process Mining's status shift

RPA's longtime sidekick — Process Mining (Celonis, UiPath Process Mining, Microsoft Minit, Apromore) — is also fusing with LLMs. Direction: analyze logs, then auto-suggest automations.

4. The code-first RPA revival

Sema4.ai (formerly Robocorp) and friends — "RPA in Python code" — are returning to spotlight. A backlash to the maintenance pain of no-code RPA.

5. Vertical-Specific RPA

Vertical-specialized RPA SaaS (finance, healthcare, insurance) is increasing. Like WorkFusion specializing in AML, the combo of domain data + LLM + RPA wins.

6. RPA-as-a-Service

Rather than building bots directly, "pay-only-for-usage" SaaS models. Zapier and Make.com already went there; traditional RPA is following.

Five-year forecast: By around 2030, "RPA" the word will likely be used less. In its place: "Agentic Automation," "Intelligent Process Automation," "AI Workforce." But the underlying activity of bots manipulating screens isn't going away.

20. Which Platform to Choose

Recommendations by situation.

Large enterprise, security/governance first → UiPath or Blue Prism On-prem deployment, sophisticated RBAC, audit logs. UiPath wins on feature breadth; Blue Prism on stability.

Microsoft ecosystem → Power Automate If you already have M365 licensing, the marginal cost is almost zero. Azure AI integration is native. Premium at $15 per user per month is overwhelming.

SMB, few bots → Power Automate or Make.com If you can't afford $10K+ in annual licensing, the per-robot model is too expensive.

Document-processing-first → Tungsten or ABBYY + any RPA When IDP is the core value, pair a dedicated IDP vendor with lightweight RPA.

Startup, developer-first → Sema4.ai (Robocorp), Browser Use, Playwright Faster and cheaper to write in code.

Korean large enterprise, Samsung/LG groups → Brity RPA or Auto Roman Domestic SI partnerships and after-sales support are battle-tested.

Japan expansion → WinActor or UiPath Japanese market share and SI network matter.

Latin America → Rocketbot Regional fit and pricing advantage.

Future-oriented, AI-agent-first → Anthropic Computer Use API + Playwright If you're betting on a 3–5 year horizon, code + agents is the bet.

Final advice: Before "which RPA is best?", ask "is RPA actually the answer in our company, or is it SaaS, API integration, or an AI agent?" No guarantee that 2018's answer is still right in 2026.

21. References

Major vendor sites

Python / open-source automation

AI Agent / Computer Use

IDP / Document Understanding

Korean / Japanese RPA

Market analysis

  • Forrester Wave: Robotic Process Automation Q1 2024
  • Gartner Magic Quadrant for Robotic Process Automation 2023
  • Gartner Hype Cycle for Artificial Intelligence 2024
  • IDC Worldwide Intelligent Process Automation Software Forecast

Technical guides / conferences

Epilogue — Bots Don't Die, They Just Change

One-line takeaway: "RPA" as a word is fading, but the underlying job of "automating the screen clicks and data shuffling humans used to do" isn't going anywhere. Whether the implementation is rule-based bots, LLM agents, or a hybrid of both — that just rotates every five years.

In 2018, "RPA adoption" meant "automate Excel and email." In 2023, it meant "automate ERP and SAP." In 2026, it means "let an agent look at the screen and decide." Same job, different tool.

Next post candidates: Replacing RPA with AI Agents — A Practical Guide, IDP Deep Dive Comparison, Korean and Japanese RPA Adoption Case Studies.

"Bots don't die. The tools change. But when the tools change, the way we work changes too."

— RPA Platforms 2026, end.

Discuss on Twitter • View on GitHub