RPA software, by process complexity and bot deployment model

What to look for, and what to ignore

What RPA Software actually costs

RPA pricing is notoriously complex. The three dominant vendors (UiPath, Automation Anywhere, Microsoft) each use different pricing models, making direct comparison difficult. UiPath charges per 'automation unit' (a bundled capacity metric), Automation Anywhere uses a per-bot-runner model, and Microsoft Power Automate charges per-user or per-flow. On top of licensing, organizations must budget for bot infrastructure (virtual machines), developer time, and ongoing maintenance. The total cost of ownership for an enterprise RPA program is typically 3–5x the license cost alone.

For a department-level pilot (5–10 bots), budget $50,000–$150,000 for the first year including licensing, infrastructure, and development. For an enterprise program (50–200 bots), budget $300,000–$1,000,000+ annually including the COE team, platform licensing, infrastructure, and maintenance. Microsoft Power Automate is the most cost-effective entry point for organizations already on Microsoft 365. UiPath and Automation Anywhere are the enterprise leaders but require significantly higher investment. If your total automation budget is under $50,000, explore low-code platforms (Zapier, Make, Power Automate) before committing to enterprise RPA.

How to compare them

Bot design and development experience

The development experience determines who can build bots and how long it takes. A platform that requires professional developers limits scalability. A platform that enables citizen developers but lacks enterprise features creates governance problems.

Ask: Does the platform support both low-code visual design and professional developer IDEs? How long does it take to build a bot for a typical invoice processing workflow? Can non-technical users build simple attended automations?

Application integration depth

Pre-built connectors for SAP, Salesforce, Oracle, mainframe terminals, Citrix, and web applications reduce development time dramatically. Without them, every interaction relies on generic UI selectors that are more brittle.

Ask: How many pre-built application connectors are available? Do connectors use APIs where available, or only UI interaction? How does the platform handle Citrix and virtual desktop environments?

Orchestration and enterprise governance

Managing 50+ bots requires centralized scheduling, monitoring, credential management, RBAC, and lifecycle management. Without these, bot management becomes a full-time job that erodes ROI.

Ask: Does the platform include a centralized orchestrator? Can bots be deployed, monitored, and updated from a single console? Is there built-in credential vault integration? Can you enforce approval workflows for bot deployment to production?

Scalability and infrastructure flexibility

As bot count grows, infrastructure needs grow proportionally. Platforms that require dedicated VMs per bot scale differently from those that support containerized or cloud-native bot execution.

Ask: Can bots run in containers or serverless environments, or do they require dedicated Windows VMs? What is the maximum concurrent bot execution on your orchestrator? Can the platform auto-scale bot infrastructure based on workload?

AI and document processing capabilities

Document-heavy processes (invoices, contracts, forms) are among the highest-ROI RPA use cases. The quality of the platform's AI document understanding directly impacts how many of these processes you can automate.

Ask: Does the platform include native AI document processing, or does it require a separate product? How does it handle varying document layouts? What is the accuracy rate for invoice data extraction out of the box?

Total cost of ownership including maintenance

License cost is 30–40% of total RPA cost. Infrastructure, development, maintenance, and COE staff are the majority. Vendors rarely include these in their ROI projections.

Ask: What is the annual maintenance effort for a typical bot? What happens when a target application updates its UI — how are bots updated? What is the average lifecycle of a bot before it needs major rework?

Before you buy

1. 1Conduct a structured process assessment: Identify 10–20 candidate processes and evaluate each on volume, standardization, stability, and rule-based logic. Score and rank them before selecting a platform.
2. 2Determine whether your processes need UI automation or API integration: If most target applications have APIs, evaluate iPaaS tools (Zapier, Make, Workato) before committing to enterprise RPA.
3. 3Define your deployment model: attended (user-assisted), unattended (fully autonomous), or both. This determines infrastructure requirements and licensing costs.
4. 4Check whether you already have Microsoft 365 licenses: Power Automate is included in many M365 plans and may cover your needs without additional RPA investment.
5. 5Budget for total cost of ownership, not just licenses: Include infrastructure (VMs), development (40–320 hours per bot), maintenance (20–30% annually), training, and COE headcount.
6. 6Plan for a proof of value before enterprise commitment: Select 3–5 processes, build bots, and measure actual ROI for 90 days before expanding.
7. 7Evaluate the developer experience with your own team: Have your developers (not the vendor's team) build a bot during the POC. Assess learning curve, debugging tools, and documentation quality.
8. 8Assess governance and security requirements: Verify credential vault integration, RBAC, audit logging, and change management capabilities. These are non-negotiable for regulated industries.
9. 9Test exception handling: During the POC, intentionally introduce variations (data format changes, missing fields, application errors) and evaluate how the bot handles them.
10. 10Negotiate contract flexibility: Avoid multi-year lock-ins until you have validated ROI. Negotiate per-bot pricing that allows you to scale up and down based on actual usage.
11. 11Plan the COE structure: Define who builds bots, who maintains them, who governs the program, and how automation requests are prioritized. This organizational decision is as important as the technology choice.