Time for a truce in IT’s battle with RPA

Scaling up is hard to do

The problems in working with the IT department could go some way to explaining organisations’ difficulties in scaling RPA. Despite the increasing popularity in using the method to cut out manual processes, just 4% of organisations were operating more than 50 robots in 2018, a negligible increase from 3% in 2017, research from Deloitte has found. Some 27% of organisations are either piloting RPA with under 10 robots or have moved into full implementation with between 10 and 50 robots.

The study also found 32% said the main barrier to achieving scale was process fragmentation and the wide variation of offline and online tasks. But lack of a clear RPA vision (17%) and lack of IT readiness (17%) were also factors.

Justin Watson, robotics and cognitive automation lead at Deloitte, says: “Few organisations have been able to scale robotics quickly, with many struggling to move beyond early experiments. Workforce behaviour needs to change to recognise the benefits of robotics and the potential boost to productivity.”

An Accenture report shows early RPA projects often make the mistake of believing they can progress without IT becoming involved because robotics tools are non-invasive, requiring no integration to legacy applications and can be installed on any desktop.

In turn, IT teams have strong negative preconceptions about RPA, assuming it was built from unsupported macros and “screen scrapers”. RPA is also unnecessary, they say, because of IT’s investment in business process management software (BPMS).

James Hall, managing director of Accenture’s automation engineering services operations, says IT often sees RPA as an inherently inelegant approach which can expose the business to risks.

“RPA is seen as an untidy way of solving the problem,” he says. “If you can enhance the application or put in an API [application programming interface], that’s cleaner and is the way you should do things. IT has seen business teams do lots of things on the desktop: rogue applications, Microsoft Excel macros and all sorts of unsupported software. If the person responsible leaves, and the technology breaks, who fixes it?”

Cooperation needed

Although most RPA projects emerge from business units or centres of excellence, they still need the cooperation of the IT department to get off the ground.

“You need IT to create robot logins for applications,” says Hall. “Usually, most login requests come through HR, so you must break with the norms to deal with that. If software teams change the underlying application, robots built to use it can fall down. To run in a scalable way, robots need to be hosted properly. They need the right server, virtual machines, permissions and so on. It all needs to be set up properly. Once you can scale, it is easier to pick and choose automation projects more effectively.”

Although many IT departments may have reservations about getting involved with RPA, there are sound tactical reasons to do so. RPA can take pressure off IT staff in situations where they face continuous demands to upgrade enterprise applications, says Hall.

“The business can get on with smaller enhancements without making requests to change the application. The more forward-looking IT departments that have embraced RPA see it as a logical thing,” he adds.

Hall says IT teams should see how RPA plans fit with their own application upgrade cycles and consider whether it could remove the pressure for an upgrade or impede an existing upgrade path. IT should also have a role in the RPA decision-making process to see if there are alternatives to desktop automation.

Bypassing IT

Lee Beardmore, chief technology officer of Capgemini’s global business services, says part of the appeal of RPA is in its promise to bypass IT. Business people look at it as a way to get a quick win without consulting the IT department.

But that may not be a wise path to take. “That’s one of the big learnings right now. RPA is not an excuse for bypassing corporate IT and its rigour,” he says.

“Yes, they can be delivered quickly, but organisations have rapidly reached a ceiling when they try to do it at scale. You have to think in terms of corporate IT discipline of traditional delivery of systems: all of the stuff associated with robust implementation. The initial push is driven by the lines of business, but they reach a point where they realise they can’t progress.”

The problem is that efforts at RPA led solely by the business units tend to miss the wider opportunity automation may present, he says. If organisations only automate based on existing application interfaces, for example, they will fail to streamline or transform processes while they are automating them.

“The way that processes are set up, even if they are streamlined in apps, is they are still quite human-centric,” says Beardmore. “That does not support a simple deployment of robotic solutions. That means when it comes to changing processes, you have to think about automation first: whether that is RPA or artificial intelligence or machine learning. It is a different mentality and way of working.”

RPA bots are good at processing big volumes of data, but not at applying subjective rules based on certain criteria, so businesses often need to simplify processes to optimise the benefits of RPA.

But that requires a more detailed knowledge of their own processes than many businesses possess, says Beardmore. “When building a robot, you have to know all the screens, all the fields, the keystrokes, and the other things that humans use to complete the task. The level of detail is sometimes underestimated,” he says.

Process mining

To understand their processes, organisations can benefit from process mining – a technique which applies machine learning to large volumes of application data to try to describe current processes and how they may be improved. But it goes beyond RPA.

“If you make it work well, it gives you great visibility about all the alternate flows that exist in the enterprise that you may not be aware of: workarounds on workarounds on workarounds,” says Beardmore. “Then you can prioritise opportunities and assess whether the policy needs to change, or if ERP [enterprise resource planning] optimisation is in order; or, if that is too expensive and takes too long, RPA is the right answer.”

There are good reasons for IT departments to be wary of business-led RPA development, but RPA can help IT by taking the pressure off the application upgrade cycle. It is also part of a group of tools to help improve process efficiency, ultimately boosting return on investment from business applications.