By Adam Devine, vice president of marketing, WorkFusion
In general, the decision-making process regarding enterprise software has been difficult – and made more so by the inflated promises of slick marketing. But business process automation has proven difficult in particular, and for three particular reasons.
The first is that providers over-sold the capabilities of the first generation of process automation solutions. Second, it’s problematic to do an apples-to-apples comparison of automation products because no two are exactly alike.
Finally, since most enterprise operations are in the early phase of automation adoption, case studies are hard to find. However, enterprises are finding ways to sort the wheat from the chaff and gladly adopting this technology. Below is a more detailed discussion of these challenges, along with three steps to help in that sorting process.
False marketing promises
Responsible providers lay out the factual capabilities of their product, along with insights learned from customers, coated in a mercifully thin layer of varnish to catch the eye of prospects. Bad marketing is just the varnish, and that’s what’s covered much of the automation market.
Robotic process automation (RPA) vendors have oversold the capabilities of their products with shiny, bold marketing, and some even market their products as artificial intelligence (AI) solutions when their technology provides strictly rules-based robotics.
Though it’s taken about ten years for rules-based automation software to find its place in the enterprise software stack, it has achieved mainstream awareness as RPA.
It is an excellent solution for automating routine, binary tasks that human workers perform on legacy applications – like entering passwords into and operating the user interfaces of SAP and Oracle – and moving structured data from one system to another.
Given the waning benefits of labor arbitrage and mounting pressure to reduce costs, the BFSI industry in particular blazed a trail into RPA, focusing on common horizontal business processes like eInvoice processing, Procure to Pay (P2P), Record to Report (R2R) and vertical challenges like KYC, AML, settlements and claims processing. In other words, the labor-intensive middle and back-office work that has for decades been mostly offshored.
Sadly, customers look back after running a POC or even a full deployment and wonder, “Where is the 90 percent cost reduction I was promised? What about the suffocating volume of manual, unstructured data work in the rest of the business process? How do I affordably handle the exceptions in the process that need human action? How do I automate the rest of the process?” This is a bit of a paraphrase, but it’s a fair synthesis of how end users react after using RPA-only products.
The problem with RPA as a category is not the technology. It’s how vendors have marketed their RPA products.
It is true that RPA automates the operation of desktop application user interfaces, and one “bot” can deliver about 1.5x the productivity of one worker, provided the work only involves structured data. It’s “hand” work – tasks that a human can perform without thinking, performed in accordance with a strict and rigid set of rules. RPA does not process unstructured data (PDFs, docs, email messages, news feeds, web content etc), nor does it have human-in-the-loop exceptions processing for when the rules governing a “bot” change.
Additionally, some RPA solutions are not deployed at a server level, which, for customers who value security, means occupying a desktop and exposing passwords and other sensitive data to human workers.
When applied to structured, rules-based tasks, RPA is a powerful feature that delivers great results, but it belongs as a feature in a complete suite of automation capabilities, and it must be paired with machine learning to automate a complex process from end to end.
The competition for dominance has begun
The various players in the market started with different theses, built different products based on their origins, and marketed to different sizes and types of businesses. Automation is halfway through the three classic software market phases: develop, compete and dominate. Let’s looks at these phases through the lens of a well-understood market: social networking.
This is a tale of three platforms. The first is Friendster, which began as a platform for creating digital ties between people who already knew one another. MySpace started with this functionality, along with increased discoverability between strangers and, interestingly, music. Facebook started as a way for students at elite universities to poke one another.
We all know how this played out. Friendster and MySpace didn’t aggressively evolve their products out of initial development during the competition phase, and Facebook became the dominant player by developing a wider range of user-friendly capability for a wider range of users.
Automation is now sitting squarely in the second phase, competition. Some of the early players have already cashed out, other players are marketing capabilities they do not have, and still others are only just now coming to market. Buyers find it hard to compare them because there isn’t a standard set of features to shop for, and some buyers don’t yet know what they need.
After all, only a few years ago, offshoring was state of the art and business process automation was an expensive, uncertain scripting endeavour imposed upon data science and IT teams.
At some point soon, the market will mature and enter the dominance phase. There will be only one scalable, widely deployed and battle-proven model, and there will be only one (maybe two) vendors who competed well and took the dominant position.
Not much to go on
Just as Amazon reviews help shoppers make purchasing decision and Goodreads reviews help readers find their next book, relatable and quantitative case studies from seasoned customers make it easy for enterprise software buyers to select a technology. These case studies are only just now surfacing, and that’s made it more challenging for buyers to make decisions.
Enterprises are still sometimes confused by what is a feature and what is a benefit of robotics, cognitive automation and AI. Because business process automation is new, there aren’t many real, believable case studies to help inform new buyers.
Most buyers are only a year or two into their business process automation journey, few of them will continue to use the platforms they started with, and many companies have only just begun their due diligence. Making meaningful bets is difficult without hard data.
Three action steps for successful decision making
First, enterprises can create centers of excellence (COEs) that centralize control of the buying process. The fastest and most efficient purchase and deployment efforts have leveraged this model, which brings together operational requirements from user groups across different divisions, product knowledge and decision-making liberties.
These COEs typically begin with reports and briefings on smart automation and digital operations from Everest, Gartner, Forrester, HfS and other leading analyst firms that are shedding light on the market. The COEs create a long list that becomes a short list that becomes a pragmatic, informed product selection.
Second, work with one or more vendors to determine which business processes to focus on to create a meaningful proof of concept (POC). A good vendor will be able to help you select the right processes, which should represent the way the business operates. If you’re a global banking or insurance operation, thousands of people ingest and process a combination of both structured and unstructured data and leverage and feed dozens of systems.
You have some legacy technology that you do not wish to disrupt, and you have some point tools that you wish to rationalize out. You have both internal and external demands to accelerate transaction times and reduce manual work while improving accuracy. You are under the gun to cut costs immediately, and any solution you consider must pay for itself in under a year. So, a good set of processes for a POC will take these factors and challenges into account.
Third, implement that POC. The CIO of one of Europe’s biggest banks once declared that if a technology could not demonstrate results in one quarter, it had no place in the operation. This is an excellent rule for automation.
The proof of concept should have literal executive buy-in. POCs are often affordable enough for divisions within companies to execute without executive support, but full deployments across an enterprise are not. There’s no point to doing a POC if your organization is not committed to modernization or transformation at the executive level. Get buy-in early.
The POC is not only a chance to see how a product performs, but it’s a chance to see how a product would deploy. Does the vendor deploy directly? Through partners? As an on-premises solution? Cloud? Desktop or server, or both? POCs are your opportunity to not just kick the tires but to drive the car, and you should drive the car hard and fast in many conditions before you buy it.
The full suite approach
Technology is the means by which enterprise operations professionals will improve margins during a time when they are down. But ops teams need to make sure they are getting straight talk from automation software vendors, not just hype. This will enable them to make informed choices that return ROI quickly. RPA is a useful feature, but it isn’t a standalone if enterprises want to transform and evolve their businesses. Rather, a suite of capabilities is in order, including business process management (BPM), RPA and AI-powered cognitive automation. What’s more, non-techie business staff need to be able to use them.
About the author: Adam Devine leads market development, product and brand marketing and strategic partnerships at WorkFusion. He began his career in management consulting in the Financial Institutions Group at BearingPoint and has spent the past 14 years in tech product marketing and advertising. He was most recently director of strategy at 360i. Adam holds a bachelor degree from the University of Vermont.