Robotic Process Automation – A Quick View
Not as straightforward as some like to think…
Many friends and colleagues know that I am working on RPA projects these days and I get a lot of questions from the uninitiated or the newly aware observers. Questions tend to follow certain themes, “Do you have to be programmer to create robots? Could I learn how to do it?” “What does it cost?” “Are robots AIs?” “Is one RPA platform better than the others?”
Of course there are no black and white answers to any of these questions. The variables that affect your approach are all about context: your own application development experience, your vision, your company’s appetite for or ambition to use these technologies, the strength of the executive sponsorship, an allocated budget, IT support to some degree (infrastructure), a clear roadmap (is it just RPA or is there an intention to grow into ML, AI, NLP, chatbots, etc).
Here are the things I think about around these particular themes:
Do you have to be programmer to create robots? Could I learn how to do it?
This question implies that a business user may want to take the lead on developing robots. Who is going to do the “programming” is a fundamental question to your RPA approach. Do you have a centralised COE or some kind of a federated model? Is it bimodal with support from IT for infrastructure? What kind of governance team are you planning?
I have had hands on experience with Blue Prism and UiPath. When I was first researching a few years ago, I used to ask this type of question and some users would respond with UiPath is more “scripty” than Blue Prism. I understand what they mean now. If you have any previous experience with programming in general, but particularly with VB or .NET, you will find UiPath easy enough. Blue Prism is definitely less “coding or scripting” oriented but it doesn’t mean you are absolved from basic knowledge of screen elements, or escape sequences, or tags, etc. I do think that a strong business analyst with exposure to macros in Excel for example, would be capable of using either but might find Blue Prism a little less daunting. It will depend to a degree on a person’s willingness to search for solutions to common issues and extrapolate the answers for their own requirements. If you are adept at that skill, you will be fine with either platform.
My own experiences also indicate that the closer the developer is to the business, the quicker positive impacts will be realised. Business process knowledge is vital to maximise options for these technologies. It’s one thing to have an analyst study a process, document it and then hand it off to a developer. That approach will work but may require many iterations between the developer and the Subject Matter Experts (SMEs). A developer who comes from the business will be able to find workarounds to issues and design solution approaches that will take into account any potential options in the process in a much more informed and creative way, than a pure developer. The reduction in development effort can be more than twice as fast.
This does beg a question of IT support. If IT is not bought into your RPA programme, regardless of your approach, it will be tough going. You are going to need them to possibly to set up a server (Blue Prism), to build you virtual machines to run your robots on unattended (budget costs?), set up target system id’s for your bots, sign off on an architecture and security review, etc. They can bog you down and become an obstacle to overcome. The best strategy is to engage a senior IT leader early, get buy in and support, and ideally a programme champion that will help you to keep momentum and remove barriers.
What does it cost?
The question of cost is bigger than just the robot licenses but let’s start there. This is an area when you really need to do your research and understand the base commitment. Some vendors require you buy license packs and commit for years (Blue Prism). Some vendors allow you to try before you buy (UiPath). And some vendors are offering to give RPA away for free (Workfusion) as a hook to get you into their wider product offering.
When you research the major players and look at the various magic quadrants, you will see some of them are clearly well established with big blue chip customers. They are still having success selling to global companies and are not yet aiming product offerings at small to medium business. I think that landscape will change soon. From a cost of entry perspective, I think UiPath is the most accessible. UiPath gives away their development tools for free and you can take their on line Learning Academy course for free and become certified as a developer and an administrator. You can build working robots that run locally on your desktop with no license fees required. This is quite an elegant “go to market” approach to build up a pool of trained developers and to access the SMB market. That doesn’t mean the costs are insignificant once you are ready to move to enterprise deployments. The cost of the UiPath Orchestrator is significant for an SMB but may still be worth it if the robots are delivering one or more FTE savings. UiPath aslo licenses single robots vs having to buy multiple license packs. You can build up to a volume of automation and manage your budget accordingly. At the end of the day, you may find the costs are comparable once you get to critical mass, but this phased approached to the cash outflows may make it more palatable.
Once the licenses are dealt with, you need to consider your internal infrastructure costs. The best practice recommendation for unattended robots is to run them on Virtual Machines. Your IT department will have some costs to pass on relative to VM licenses and possible other related overheads. If you need a server, that is an added costs also. Between the hardware and the set up fees you might be charged, this could be a significant portion of your initial project costs. Your ROI calculations will take into account initial set up costs with lower payback in year one. The big pay back comes in the subsequent years as you increase the utilization of your robots
Are robots AIs?
This answer is NO. A Robotic Process Automation, is a rules based, structured data process. There is no thinking, learning, qualitative decisions, emotion, sentiment… It is nothing more than a macro through the UI of the applications you are using. This type of question though, leads to discussions on a roadmap where you may be considering handing off data processed by a robot, to a system that is learning, making recommendations, classifying data, performing some kind of analysis, etc. These kinds of technologies are a natural extension to RPA but not required to be successful (from a payback perspective) with RPA.
Is one RPA platform better than the others?
I have seen countless demo’s of a various RPA solutions. They are all great. Each one may have some benefit or cost compared to any other one. The biggest mistake is to not get involved or delay your involvement with these technologies because you can’t decide which one to use. All of them will pay back their costs plus more, provided you have found suitable applications.
I see RPA as an easy first step towards a promised nirvana of AI’s all things autonomous and intelligent. It is still the least expensive technology to start with compared to machine learning and AI’s. Some of you are probably already familiar with open source tools for machine learning, performing various image recognition tasks or classification tasks for example. While these base levels of functionality are essentially free, the costs to those tools quickly rise when you want to extend those capabilities specific to your business. The cost to add the required labelled images of your industry specific products for example, to an open source tool for image recognition can be significant. The commoditization of ML and AI functions is coming. By the time you master RPA, these tools may be accessible to all size businesses but in the meantime, get on the RPA train and start your journey…