kirill_makarov -

A bottom-up and top-down look at intelligent automation

We look at robotic process automation at a coding level and an organisational level, and see how artificial intelligence fits

There is a misconception about robotic process automation (RPA) that bots will replace jobs. This is not true. Bots have very little in the way of intelligence. The limited level of decision-making that can be automated involves simple if-then decisions – “If I see this, then I need to do that” – which are at the heart of computer languages. 

It is true that a bot can indeed read some data on an input screen, make a simple decision and then perform some action on the input screen of another application. People do this all the time, but it does not require a huge amount of brain capacity.

Moving data from one place to another, based on a set of simple criteria, is not exactly taxing. But if this simple task needs to be repeated multiple times a day by many, many people across an organisation, using a bot can lead to a big boost in productivity.

The reason bots are needed is often that enterprise applications are deployed to solve particular business needs. Sometimes, too little thought is given to how a new application fits alongside existing software.

It doesn’t make sense for someone to rekey information between systems, but that option may be far easier for the business to accept than the prospect of having to build custom integrations between enterprise systems using application programming interfaces (APIs). The RPA replaces the task of manual rekeying.

Automating invoice processing

Mathieu Webster is transformation lead for the finance and accounting function at NHS Shared Business Services (SBS). He is part of the NHS procurement shared service team, responsible for handling more than seven million accounts-payable transactions, managing £225bn in cash and recovering £20bn of debt for the NHS every year.

Webster says automation and scanning play a huge role in getting invoices into the Oracle system that handles NHS procurement.

About 850 separate financial processes are carried out by NHS SBS, including reconciliations, cash flow, invoice payment and debt collection. At the beginning of its RPA journey in 2016, the organisation identified that almost half of these were suited to automation. It has since completed the development and deployment of an entire ecosystem of robots handling more than 250 of these processes.

Describing a typical automation run through RPA, Webster says: “We have to match purchase orders with suppliers.”

Suppliers may operate from several different sites and deliver across the NHS. UiPath has been used to streamline the procurement process by linking different applications to create a workflow that requires little or no human intervention.

“It’s screen-scraping,” says Webster. “We check that the logo on the invoice matches the supplier’s logo and we use decision flows to decide whether the invoice can be fast-tracked into Oracle, whether we need human intervention, or if it needs to be rejected.”

NHS SBS uses screen-scraping to copy and paste information from one application screen into another, controlled via a bot. This effectively replicates the actions of a human operator. Direct access from one system to another, via an API, may seem a more efficient approach, but in Webster’s experience, it isn’t always the best choice when organisations need to link disparate IT systems together.

“In some cases, systems we don’t have API access to, such as the Oracle system, didn’t have an API we could use,” he says. Apart from requesting and then waiting for an update that includes the necessary API from Oracle, the bot can do the work without requiring any developer work, says Webster. “You can follow a standard business process to input whatever the system needs and it will work as expected.”

Unlike an API, which involves creating new code that interfaces between systems, the bot is simply doing the same actions as a human, which reduces the checks and balances that need to be done if the same automation was achieved using new software integration, says Webster.

Because UiPath supports Amazon Web Services (AWS) and Microsoft Azure cloud-based machine learning, NHS SBS has been looking at how to build intelligence into the bots. One example is a proof of concept workflow analytics application in employment services.

“We take data from people who are leaving and use machine learning to look at who is likely to leave,” says Webster. The raw data is processed using artificial intelligence (AI), which then kicks off a process through RPA.

Bottom-up approach to automation

Danilo McGarry is global head of automation and AI at Alter Domus, a specialist in the alternative investment industry. McGarry has worked on major RPA initiatives at UnitedHealth Group and the Royal Bank of Canada (RBC), where, over six years, he was involved in projects that automated a significant portion of RBC’s total capital markets workforce

Describing his role at Alter Domus, McGarry says his “mission was to set up a brand new automation department”. From his experience in running RPA projects, he regards automation as a culture shift.

As for why Alter Domus needed an automation department, he says: “We had never seriously created a culture and change mindset. We are growing fast and need a backbone.” 

Supported by technology, McGarry set out to establish a centre of excellence to promote and support the wide-scale roll-out of automation to tackle highly repetitive, low-value tasks. “It’s a dream to go into a company and do this from scratch,” he says, which has allowed him to build a team comprising the best developers in RPA and internal business analysts.

But there is more to successful automation than a centre of excellence. For McGarry, automation is a transformation journey. “Year one is slightly different to year two and three,” he says.

The most important criterion that determines how successful an RPA project can be is the people, says McGarry. “Make sure [any RPA initiative] is all-inclusive from the ground up, where people have the ideas we can work on together,” he says. “We empower them to give us ideas.”

McGarry is a big fan of running workshops, which introduce employees to RPA and help them understand how it fits into their work. Through the workshops, it is possible to identify people who are genuinely interested in how RPA can help them do their jobs more effectively. Such people are encouraged to develop RPA skills, such as UiPath Academy training, which results in industry-recognised certification.

McGarry says that in his experience, this can help to boost people’s future job prospects. “We hand-hold during a three-month programme, then they can do their own projects or make one of the bots on our to-do list,” he says.

There will be times when people find that a task they have done for years is being automated. This leads to anxiety that an RPA initiative will reduce their job role, make them less valued or potentially lead to redundancy.

McGarry’s cheat sheet for helping employees make transformative changes involves one-on-one conversations. “I sit with people and ask semi-life-changing questions like, ‘What you think of your job?’, ‘How often do you work in automatic mode?’ and ‘How often do you enjoy coming in to work?’ More often, it’s for the pay cheque,” he says.

But he believes this is not necessarily the life people want. 

Read more about bots and AI

  • RPA and AI adoption are increasing rapidly, but there is still some confusion over the co-evolving technologies. Here we dive into their capabilities and how they can work together.
  • Instead of an all-or-nothing approach to supply chain automation, RPA is most effective when targeted at sub-processes to improve high-volume, repetitive, error-prone tasks.

People often believe that bots are really not good at doing complex things, says McGarry, and when people question him, he has a standard answer: “This is not true. You have a beautiful human brain. It’s the most complex machine on Earth. You should use that to do interesting things. Let’s get you to be human again and take the robot out of you.”

For McGarry, such conversations are a bit like going in as a coach in a third league football team. “We should be a premier team,” he says. “I inspire them to do something new, give them a more interesting job and learn new skills.”

Another tip in McGarry’s arsenal of tactics to sell the idea of RPA to employees is his analogy of Iron Man. “Do you want to be a superhero?” he says. Iron Man is just a normal person with an extraordinary armoured suit. In effect, RPA is a kind of armoured suit that enables employees to achieve superhuman work.

He urges business leaders not to try to achieve a payback on an RPA initiative in year one. In McGarry’s experience, the first six months to a year is spent setting up the RPA team.

“If you try to get a return on investment in year one, you become more ruthless,” he says. This then makes the RPA programme seem less about employee productivity and more about business efficiency.

Rather than trying to achieve a return on investment, McGarry recommends that organisations focus on quick wins. “Robotic process automation works most effectively when an employee gives a testimonial to the business,” he says. “It spreads like fire, offering people an opportunity to get a better work-life balance.”

Year two is when an RPA programme shifts into results mode, says McGarry. This involves running workshops and talking to people who run departments to build a “to-do” book of ideas for automation. By year three, he says organisations can start to consider techniques such as process mining and behavioural and time analysis studies, which can help to build a heat map of which tasks people spend most of their time doing.

Removing integration blockers

No business application is ever truly standalone. At some point, information needs to be shared with other enterprise systems to join up business processes. However, a recent survey of 800 IT leaders conducted by Vanson Bourne for the Mulesoft 2021 connectivity benchmark report found that almost nine in 10 respondents point to integration challenges as a blocker to delivering on digital transformation.

Processes that are disjointed require human intervention, preventing different workflows from being handed over seamlessly. APIs offer one approach to unblocking digital transformation challenges, but as NHS SBS found out with its Oracle integration, such APIs may not always be available. A bot can often provide a much easier path to integration.

Another recent study by Knowledge Capital Partners for RPA provider Blue Prism discussed why RPA is part of a much bigger picture, as an enterprise platform integrating other technologies to enable digital transformation.

For McGarry, RPA can’t be used to do everything. It is part of a toolset that organisations draw on to transform digitally. This transformation may require developer work and the use of APIs. But RPA, AI-based optical character and image recognition for reading scanned-in paper documents and AI-powered natural language processing to understand customer correspondence all have roles to play – as does the use of AI to automate some of the decision-making steps in a business process. But these are smaller parts of the bigger picture.

Arguably, the most important tool in the toolbox is the human element. Giving people the confidence to develop their own ideas of how to improve work processes, and providing them with support to automate inefficient tasks, is perhaps the most intelligent use of intelligent robotic process automation.

Read more on Artificial intelligence, automation and robotics

Data Center
Data Management