RPA Vs. Process Mining: What Companies Should Know


Those who follow trend lines in enterprise technology have seen these two things bandied about at increasing rates for the last few years. Technologists and executives who have worked companies that have leveraged RPA and process mining may understand the differences. For most others, however, the subject is likely more opaque.

Consider this a guide to understanding how RPA and process mining both work and how each of them is different. For CIOs and others helming companies’ software decisions, this should be required knowledge. Investors, too, should be tuned in here, as UiPath hit a $10.2 billion valuation with a $225 million raise in July 2020, and Celonis’ valuation was $2.5 billion in its $290 million raise in November 2019.

Both companies have spawned an army of copycats pouring out of Silicon Valley and other tech hubs. One VC I spoke with commented that many of these outfits are rarely the software platforms they claim to be, but really just a couple of engineers writing scripts to automate clients’ badly implemented solutions.

What is RPA?

RPA works by simply enabling automation within applications and interfaces that people already use. For example, an internal accountant who is toggling back and forth from Great Plains to a warehouse management system, and copying items from one system to the other, or making other verifications by eye, could hand these manual processes over to RPA.

RPA works similarly to other screen-scraping methods, and it has been further enhanced by OCR (optical character recognition), which helps RPA bots work better inside software without plug-and-play integrations or consistent DOM structures in the case of browser-based SaaS.

Key to the attraction of RPA bots: their ability to automate and integrate where out-of-the-box connections or APIs don’t exist. It’s ideal inside of legacy ERPs or old custom software that is pervasive throughout large companies. Avoiding expensive implementations of new software can be quite attractive when the same kind of automation that the new software would deliver is theoretically available using an RPA bot that can bolt on to existing systems.

The idea has always been that a company can set out with a good RPA platform and automate away most of its onerous manual processes and free up its employees to do value added work. This, in theory, boosts the value of the enterprise and provides workers with better workdays and levels of job satisfaction.

While these things are certainly possible with RPA, setting the platforms up isn’t simple, often requiring fleets of expensive consultants. RPA bots can also be brittle, breaking when interfaces change or a company puts a new wrinkle into its processes, which necessitates bringing consultants back in to fix them.

The RPA space has been particularly hot with venture capital investors who focus on B2B products, as there are at least 25 (by our count) RPA startups who have drawn significant funding during the last few years. UiPath remains king here.

Startups in the space have looked at UiPath’s architecture and some of its tech choices—bots are written in either a proprietary UiPath language or VB.NET—and thought they could do better. It’s the natural cycle of things in software, of course, in that each new company hits upon something it claims will separate it from the growing pack.

There also remain rafts of RPA skeptics, technologists by and large, who see it as nothing more than a band-aid for deeper technology issues and a lack of overall integration. Build out better tech footprints and perform better implementations in the first place, they say, and the need for RPA goes away.

The other issue with RPA at sprawling enterprises is that they don’t know where to use it, in many cases. Many manual processes may be hidden out of site to managers and technology leaders at companies, which, obviously, makes them hard to automate.

This is where process mining comes in.

What is process mining?

In large companies that run many dozens of software platforms, tracking processes end-to-end can require legions of whiteboard exercises with people drawn from every department of operations. Finding the weakest points in a company’s set of data flows can therefore prove difficult.

Process mining, however, automates much of the hard work here. This software extracts data from application logs and apps databases to paint a picture of how data is flowing through the company, and where the pain points might be, based on delays in transmission, the rate of data errors from one platform to another, and other telltale clues.

Process mining software—the largest player here is Celonis—will offer visual representations of a company’s processes and data flow while highlighting the pieces that are most inefficient.

Process mining software can help find ways to better connect disparate pieces of technology by surfacing data flows that are indirect and ripest for automation.

These process maps can be quite handy—some even offer insights around how often a given process performed via manual methods, helping companies put scrutiny against operations costing it the most.

Putting RPA together with Process Mining

Putting process mining and RPA together offers clear benefits—the first finds the places where more automation should be applied and the second brings the engine that can do the work. This led process mining outfits such as Celonis to build out sturdy integrations with RPA bots—the company has connectors to UiPath, Blue Prism, and Automate Anywhere.

On the other side, RPA companies realized their products could be greatly enhanced by process mining capabilities. To that end, UiPath purchased ProcessGold in 2019 to shore up its liabilities related to the quickly growing space occupied by Celonis.

It’s a natural combination. One method provides the prognosis, the other provides the medicine. That Celonis was the first company to hit upon this construct in a big way is rather remarkable, given that companies have been employing vast collections of applications and data sources for decades—most of which include transactional data and app logs that could inform a holistic reading on process flow.

What seems intuitive to everybody now, of course, certainly wasn’t before Celonis made a great business out of it.

Task Mining: Finding Other Work Processes Such As Email, Scheduling Tools, SMS

The other pieces of the work universe that these companies are now bringing into focus are all of the work processes that take place outside of packaged applications. Email, for one, remains a mainstay in many workflows, from submitting and receiving invoices to providing confirmations and notifications.

Software to track these less structured electronic environments has emerged. The industry refers to this as task mining. Application logs and data tables are usually not enough to paint the entire picture of what happens in these apps, so tracking processes here requires more nuanced tools.

For many companies, work inside of email and other similar platforms represents a major piece of the pie—multiple hours per day—for a good cross section of employees. Tracking this, therefore, becomes critical when trying to build out a total picture of work and processes within an enterprise.

Task mining software companies have been duly snapped up by Celonis, UiPath and other major players, as these companies seek to provide their users with complete solutions. This is an evolving part of the ecosystem, and we expect it to be one of the more active spaces in the enterprise application space.

Getting More Data Out Of ERPs and Business Apps

The persistent challenge for larger companies running enterprise-level systems such as Oracle and SAP has been getting data out of those ERPs and into other software platforms, including standard data warehouses and BI programs. That challenge remains for companies who want to leverage data from their core ERP processes inside RPA and process mining applications.

UiPath, Celonis and others are getting better at pulling data from these systems, but Oracle and SAP don’t make it easy, as they’d prefer that companies stay at work inside their application rather than bouncing out of it.

Closing The Loop To Action

The obvious end-game for any of these solutions is enabling automation to make informed decisions, and then take action, based on the data and business conditions at hand. That may mean issuing an incremental invoice to a client who only paid part of what they owe, for instance. RPA bots came about to solve these exact kinds of problems.

The issue, however, for RPA platforms is that they were designed to do one thing: automate a particular task. They’re not built to make decisions based on a raft of existing data and past decisions. They’re more about human-devised if/else statements than developing decision trees and algorithms on their own.

But that next level of automation is coming. In the future, the winning machines will make context-aware decisions based on a confluence of factors that sample data from across the enterprise. The race is on to integrate machine learning into RPA and process mining to achieve that. Companies in which mining process is the core product—such as Celonis—would seem to have an edge here, but nobody has cracked it yet.

How much do RPA and Process Mining Cost?

These are software platforms built for enterprises. They’re not cheap, and their implementation may require a small army of consultants, who cost as much or more than the software. There are few implementations of RPA platforms or process mining that don’t run well into the six figures.

We’ve seen many cases where RPA bots are being used a rough fixes in place of tighter data integrations and better ETL processes. Every company can use more smart automation, but RPA bots often sit on top disjointed software landscapes and end up hiding technical debt. And when RPA bots break, usually the first call goes out to the expensive consulting shop that first set it up.

Rather than setting off on RPA odysseys, many companies will be better off unraveling and mitigating some of their technical debt to enable straight-forward data integrations and automations inside of their existing software.


Source link

Translate »