Robotic process automation can streamline business workflows by eliminating tedious manual tasks without requiring you to completely re-engineer legacy systems.
If your organization is like most, you probably have a large chunk of old code that continues to work perfectly well, at least according to its original specs. The problem is that the original specs imagined that your business users would enjoy spending their days clicking on the same boxes in the same pattern and waiting for the same screen to refresh.
Enter robotic process automation (RPA). More a means for streamlining work with old software than anything to do with Asimov’s three laws of robotics, RPA is here to save your organization from boring, repetitive tasks better suited for a machine by adding another layer of automation to your stack that will click on those boxes for us. And it’s not just because clicking a button is such an imposition. The new robotic layer can also add more safeguards to prevent mistakes.
Some companies market RPA as “workflow automation” or “work process management.” Others distinguish RPA from plain-old “business process automation” by saying that RPA includes newer, more sophisticated artificial intelligence and machine vision routines. The tools have definitely gotten smarter at dealing with legacy screen shots and paperwork, but in the end, tools under each of these headings are all about relieving humans from the job of bossing around old software.
Interest in RPA is rising in part because organizations are realizing that, despite all their best intents, rewriting the old code means reading it, rethinking it and redesigning it — often starting with an old language few have the talent and experience to go at. So, if the software works, maybe it’s better to leave it alone and just glue another layer on top.
Plus, rewrite old code isn’t possible if it’s someone else’s. RPA bots are generally as adept at accessing data controlled by partners or, perhaps, even competitors as they are with local data. Parsing web pages and extracting the right data are common tasks for these bots regardless of whose web pages they are. All of the standard file transfer mechanisms like SFTP are old hat and the tools can also work with many of the newer ones like Dropbox and Google Drive.
If the web is like a candy store for these tools, the old green-screen interfaces may be even better targets because who is going to tweak those to make the interface look fresh, thereby breaking your bots’ routines. All you have to do with most RPA bots is let them watch you click around.
And it’s not just through the user’s screen where RPA does its magic. Machine vision routines, for instance, are increasingly able to suck numbers and words out of images of drivers licenses and other documents. These make it much faster for industries like banking and insurance to automate work that deals with physical documents.
The biggest advance, though, may be the way the tools are “programmed,” or “trained.” Most of the programming isn’t done by typing instructions in rigid syntax. Instead the robots often “learn” by watching business users click away. They then use this click stream to imitate what your users just did — similar to how spreadsheet macros can be created.
Still, RPA is not be as automatic is it sounds. The sales literature tends to downplay the amount of manual intervention and tweaking necessary during training. Do the robots jump over to Facebook to like a few photos before moving on to processing the next invoice? No. You will need to clean up some clickstreams not just to purge the detours but also to help the bots deal with ambiguity on the web pages.
In some cases this will mean writing code to handle something that can’t be done by a preconfigured bot. But you likely won’t have to do very much. The standard inputs and outputs work well and you’ll probably only need special code for weird file formats or strange interfaces. But the bots keep getting smarter, making training ever easier and edge cases less frequent. Artificial intelligence routines can also help look for patterns that may speed up the bots in the future.
If you’re ready to start welcoming robots into your workflow, here are 11 of the top RPA tools for streamlining your workflows and saving your users from the tedium of old software, as well as some open source projects to check out.
Blue Prism, one of the earliest RPA companies that began in 2012, talks of a “code-free canvas” for teaching your robots to emulate humans. The emphasis is on using artificial intelligence and machine learning to smooth the way the robots behave over time. You string together a sequence of actions at the beginning, but then each action generates statistics that can be used to train and improve the choices made. The code, if you can use that word for the actions, learns as it goes.
The company also maintains a digital exchange where plugins and add-ons from others can be purchased to extend the powers.
The robots from UiPath are designed to be installed on Windows desktops and servers where they can take over any of the tasks that might normally be done with the screen. These robots are controlled by another process, the Orchestrator, which will trigger them in response to events. UiPath is expanding rapidly into artificial intelligence and is also emphasizing machine vision tools that can extract information from images or screen shots. These are often focused on optical character recognition to convert letters and numbers into machine-understandable forms.
The Bot Store at Automation Anywhere offers a big collection of tools that will do all the standard clicking and tracking as well as some that can glue together many of the traditional complex data files that float around the internet. There are bots for extracting information from spreadsheets, files or web pages and then another set of bots that will store this information in databases for issue tracking, invoice processing and more. Many of the bots rely on outside APIs such as Microsoft Azure’s image analysis API. They also offer a “community edition” that is free for small businesses with a limited workflow. This edition can also be an ideal form for testing and exploring the technology.
IBM offers a wide range of options for automating menial tasks, split into separate products. Information enters the pipeline through IBM Data Capture, flows through paths defined by Business Automation Workflow, where it’s final resting place is decided by IBM Operational Decision Manager. The company has been refining these tools for some time because the product literature says they’re on the 25th anniversary of managing workflow.
Lately they’ve been partnering with Automation Anywhere to integrate the two stacks and use some of the latest tools on the workflows defined inside IBM’s stack. Both of these can be deployed locally or in IBM’s cloud.
ImageTech Systems makes Kofax, another set of bots that can suck data from many of the standard file types (Excel, JSON, CSV, email, etc.) and then act on it. One of the nicer features is the Robotic Synthetic API, a nod to the traditional ways of programming. Code written in Java, Python or another programming language can reach out to the bots and give them instructions, making it a bit easier for your regular stack to interact with the RPA.
The bots can also be spun off into smaller tools called Kapow Kapplets that can handle focused chores locally. All of the behavior is tracked with standard analytics and reported through a dashboard so you can watch for robotic glitches.
If your process is heavy on documents that may need signatures, Nintex’s collection of RPA bots includes tighter integration with Office365, Salesforce and Adobe tools. The users may feel like they’re working with real paper, but the work is done digitally and the flow is managed by the tool. Nintex calls these “logic-driven documents.” Of course, you can also automate standard data sources if you don’t need to produce “documents.”
There are two parts to the Kryon RPA. The first “discovers” the process by recording everything while running quietly in the background on employees’ desktops. (Some applications and web sites can be excluded for privacy.) These automatic trails can then be edited in Kryon Studio to fine-tune the actions. The resulting code can be either fully automated, as an assistant for humans, or in “hybrid” mode, which is something in between.
Kyron’s tools also include artificial intelligence for extracting numbers and letters from images and more sophisticated machine learning algorithms for optimizing the process over time.
TruBots, the name Datamatics gives to their individual programs, are created with TruBot Designer, a tool that allows you to create and edit the software. It begins by watching keystrokes and mouse clicks, but also offers an IDE for fine-tuning. Much of the work can be accomplished by dragging and dropping standard components, but developers can also adjust the system-generated code in the IDE.
The actions of all of the bots can be coordinated with the TruBot Cockpit, which will deploy and monitor their actions. The system emphasizes text processing with special tools for scanning images (TruBot OCR) and making sense of unstructured text (TruBot Neuro).
The name of the company is a Japanese term that might be translated, “automation with a human touch,” a reference to how its software robots are trained to emulate humans interacting with the standard systems (mainframe terminal, web, databases). The Jidoka dashboard tracks all of the running robots and can also create a video of the screen to help in debugging. Jidoka operates its own cloud for deploying the robots or you can move them into your own premises.
The bots from WorkFusion learn their tasks on Windows machines and then go to work on a Windows server using a mixture of repetition, optical character recognition and some artificial intelligence. The architecture is already tuned to some common challenges such as anti-money laundering, insurance claims and coverage decisions, and banking services. These can be a starting point for anyone tackling similar workflows.
The company also offers a free tool, the RPA Express Starter, that is limited to one bot running locally and doesn’t come with access to the more sophisticated machine learning bots, but it still makes a good place to experiment on easier applications.
The bots at AutomationEdge can focus on web pages, databases such as SAP, Excel spreadsheets and chatbots, in case you want to connect to customers through a chat session. The bots watch all these options and execute when new data arrives. The spreadsheet bot, for instance, can be triggered by a change as small as one line. Many of the bots available in the bot store are pre-configured for particular industries or sections of a business such as human resources or customer relations.
AutomationEdge also offers a free version, the F3 RPA Bot, that’s limited to one bot but offers a full version of the various features of the platform.
The major companies are generally selling proprietary tools although community editions with limited functionality are common. Open source processes are less common but you can often accomplish many of the simple tasks by stringing together some open source projects. In many cases, you’re going to have to do much more work to train the tools yourself, often by typing code into an editor. Still, they remain an interesting option. Check out Puppeteer, Selenium, and Headless Firefox for a basic start.
Content adapted from “Top 11 RPA tools” written by Peter Wayner
Obteined from https://www.cio.com/