RPA has long been seen as a key enabler of digital transformation. In 2021, Deloitte found that 78 percent of organizations were implementing it, with a further 16 percent planning to implement it in the next three years, and only six percent saying they had no plans to adopt.²
This near-universal uptake is due to RPA’s compelling value proposition. In the ideal scenario, it enables software bots to take on mundane, repetitive tasks, freeing people to do higher-value, more enjoyable work. RPA is fast and affordable to implement and doesn’t require back-end integration work because the automation is conducted at the user interface level.
As a result, business processes speed up, error rates plummet, and employees are more engaged. Costs also go down, revenue increases, and the customer experience improves.
View all chapters
And there’s no doubt RPA can deliver on this promise. Organizations across the world are reporting transformational benefits from delegating routine tasks to bots.
Take NBN, the Belgian agency responsible for developing and publishing standards. In 2021 it deployed a bot to take on the task of gathering and keying in ballot information—a laborious job previously done by hand. As a result, NBN went from publishing 800 standards a year to publishing 2,150. Since each newly published standard has been shown to increase Belgium’s Gross Domestic Product (GDP) by EUR 2.04m per year³, the bot has had a significant impact on the Belgian economy.
Or Primanti Brothers, the iconic Pittsburgh, PA-based restaurant chain. In its 89 years, it has never stopped innovating—and in 2021 it identified an opportunity to automate the production of daily sales reports from 40 sites using an RPA bot. A task that once took each regional manager 45 minutes per day is now done in three minutes, saving over 2,000 hours annually. The gift of time means managers can spend more time focusing on delighting the chain’s fans.
Or Credigy, a specialist Atlanta, GA-based consumer finance business. It saw the strategic potential of RPA as early as 2018, identifying repetitive, routine tasks across the business that could benefit from automation. Low-level tasks that previously took up much of its experts’ time, such as manually checking and re-naming thousands of files, were delegated to a bot. In the first year of using RPA, Credigy automated 25 processes, freeing its highly skilled analysts to work on negotiating the complex deals that underlie the creative financial solutions it provides to customers. Since then, automation has freed up even more time, helping the firm to maintain a double-digit compounded annual growth rate.
These examples show what’s possible when RPA is implemented as part of a business-led strategy that takes time to analyze processes and identify where automation can be most effective.
But for up to half of organizations that have deployed RPA, the results haven’t been so stellar.⁴ Either RPA hasn’t delivered the expected ROI at all, or the early wins haven’t scaled up into continuous, enterprise-wide optimization.
This can sometimes feel like a software problem. But in almost all cases, the problem isn’t with the RPA solution so much as the way it’s deployed.
In particular, the low barriers to entry with RPA mean it’s often implemented within a single department, to automate specific tasks that are tying up employees’ time or which seem to be causing bottlenecks and inefficiencies.
For example, imagine an organization that has been receiving complaints from suppliers about late payments. Time spent manually keying in data from paper invoices seems to be an issue, so the finance department uses RPA to build a bot that scans invoices and enters the details into the SaaS finance system.
The bot works well but procure-to-pay (P2P) lead times don’t seem to improve. Eventually, the person who built the bot leaves. Nobody else knows how to keep it up to date, so when the SaaS vendor next updates the finance system, the bot breaks.
This simple example highlights several ways RPA implementations can fail:
Only one part of the process was addressed: Procure-to-pay and order-to-cash are some of the most complex processes in the modern organization, spanning multiple departments and external stakeholders, and comprising long chains of interdependent tasks. Tackling just one part of the process may bring some relief, but if there are bottlenecks elsewhere in the process, the overall improvement may be minimal or non-existent. An isolated fix to one part of the process can even create new issues or bottlenecks upstream or downstream.
Automation was applied to a bad process: This process involved paper invoices, which may have been mailed to the organization, sorted in the mailroom, passed on to the customer’s contact, and perhaps lost for a while on a messy desk before being physically passed to Accounts. Automating data extraction from those invoices has done very little to accelerate the whole P2P cycle. The whole process needed rethinking before applying any automation.
KPIs weren’t clear: The organization leaders knew there was a problem but didn’t think through what improvement they wanted to see. Rather than specifying a desired outcome and measurable KPIs, they just applied RPA as a sticking plaster to one of the more visible sources of inefficiency. By stepping back and examining the whole process, they could have pinpointed all of the inefficiencies, decided how to fix them, and calculated the ROI of each fix before acting.
Ad-hoc tool use made automation unsustainable: The bot was built by a tech-curious finance person using a low-code RPA tool. They found the software easy to use, but that knowledge departed with them when they left. Not only did this lead to the bot breaking but it was also a missed opportunity to scale the use of RPA to tackle other inefficiencies.
Ongoing governance was lacking: Because the bot was built and deployed ad-hoc rather than as part of an automation strategy, nobody (and no monitoring system) was keeping an eye on it, so nobody could predict that an update to the finance system would break it.
What’s missing from the scenario outlined above are five elements that could have made the use of RPA successful—both in the short and long term:
Adding these elements may seem like a lot of work—and until now, it has been. Traditional Business Process Modeling (BPM) is a labor-intensive activity that can take months and involves manual mapping and analysis of organizational processes to identify where efficiencies can be achieved.
But now, there’s a much faster and better way to get insight into the hidden inefficiencies holding organizations back—and it’s already unlocking billions of dollars of untapped value every year for companies around the globe.
Enter Process Mining.
Bring visibility, governance and scalability to RPA with Process Mining.
² https://www2.deloitte.com/bg/en/pages/about-deloitte/articles/Intelligent-Automation-Survey-2021.html (link resides outside of ibm.com)
³ https://researchportal.vub.be/en/publications/the-economic-impact-of-standards-in-belgium (link resides outside of ibm.com)
⁴ https://www.ey.com/en_us/consulting/five-design-principles-to-help-build-confidence-in-rpa-implement (link resides outside of ibm.com)