Making the initial case for RPA is easy—every modern business can see the value of shifting mundane, repetitive tasks from humans to robots. RPA, or Robotic Process Automation, offers the promise of reductions in labor costs and cycle time, greater operational efficiency, and elimination of human error. This all sounds great, right?
RPA software providers have carefully crafted messaging to promote all that is bright and shiny about RPA, but stop short at explaining how difficult and expensive it will be to build it yourself. And why wouldn’t they gloss over this? Would companies purchase their multi-thousand-dollar licenses to conduct proofs of concept if they knew that 30-50% of RPA initiatives fail? Or, worse, that less than one percent of RPA initiatives reach full maturity with multiple bot deployments across all lines of business? They're in the business of selling software, not helping you make your processes work.
To be clear, when building your own program, the magical ROI of RPA being touted across the internet will only be realized when your program reaches maturity. Sure, you will see incremental lift as you start knocking out those high-touch, high-volume processes, but the savings realized early on will not surpass automation’s total cost to the business. Let’s unpack the why behind that.
HfS Research estimates that licensing costs represent only 25-30% of RPA’s total cost of ownership (TCO). The remaining 70-75% represents the part that no one is really talking about: the ongoing cost of support personnel.
To simplify the concept of TCO, we’ll divide it into two categories—cost of technology and cost of support. Within each category are one-time, upfront costs and ongoing, annually-recurring costs.
Breakdown of Technology Costs
One-time:
RPA vendor licensing
RPA vendor training
Infrastructure setup
Third-party integrations
Proof of Concept
Ongoing:
Annual RPA license(s) renewal
Annual maintenance renewal
Annual third-party integration license renewal
Infrastructure maintenance and/or expansion
Vendor management (where applicable)
Management of program SLAs
Let's zoom in on vendor licensing costs. As HfS Research points out, RPA software pricing and licensing models are notoriously confusing. The technology set is typically broken down into separately-licensed modules dedicated to a specific use—bot builders, bot orchestrators, bot dashboards, and advanced analytics hubs. License models can be annual, perpetual, consumption-based, or SaaS. Licensing costs can also vary between unattended and attended automation, number of users, number of bots, number of machines, and run-time. Some vendors even carry bot or license minimums. As if you needed additional challenges in building your own RPA program, pricing across all vendors remains unnecessarily complex thereby making it difficult for your COE to keep tight control of total technology costs as your program expands.
Breakdown of Support Costs
One-time: Establishing the COE
Hire or reallocate analysts and developers to plan, design, build, and maintain the program
Secure IT support for infrastructure and deployment
Ongoing:
Annual overhead for support personnel
Ongoing training
Continuous process optimization
Continuous expansion of the program
Continuous bot monitoring
Continuous bot maintenance (Note: The average life cycle of a bot is 18-24 months)
The RPA Center of Excellence (COE) is essentially a dedicated steering committee responsible for guiding, managing, and building your RPA initiative. This COE is typically comprised of 5% stakeholders, 25% business process experts/analysts, and 65% IT personnel. So, let's say you start small with a COE of 15 people. Depending on the provider selected and the processes being automated, your COE will require between three-to-nine developers. With an average salary range between $100K-$160K for RPA developers in the U.S., you're looking at a minimum annual cost between $300K-$480K just for IT personnel, on top of annual licensing costs. It's a significant expense that must be factored into the overall ROI of the program.
But, what about pre-built bot libraries? Won't that help reduce the need for experienced developers? Not exactly. Every major RPA vendor provides user-generated, pre-built bot libraries to "accelerate" RPA deployments. So, one would think it's a safe assumption that subject matter experts (SMEs) at any experience level could throw together an automated process in an afternoon, right? Great in theory, problematic in execution. Here's why. Most applications available on the market today are completely customizable, and few companies use applications "straight out of the box", thus edits to pre-built bots are practically guaranteed. Using proprietary or highly-specialized applications? You’re likely building from scratch every time. Automating Citrix-based applications? Prepare for extra maintenance as automation that utilizes computer vision is significantly less stable. Bot libraries can be a great start, but they in no way remove the necessity for qualified personnel.
There are myriad factors that can lengthen time-to-implementation, require more expensive or expert personnel, or increase change management (and downtime). These unforeseen factors, while difficult to calculate, can have a significant impact on the overall cost and success of your program. Understanding this from the outset is imperative to building a smarter, more scalable program.