Leslie Willcocks, Professor of Technology Work and Globalisation, London School of Economics and Political Science.
John Hindle Knowledge Capital Partners
In our most recent research we have been tracking organisations over time to see how they grew their strategic use of RPA and cognitive automation. One such company is Nielsen, a global information, data and measurement firm. Nielsen operates in over 100 countries and employs approximately 44,000 people worldwide. Total revenues were $6.515 billion in 2018. We talked to two, senior executives at Nielsen—Deborah Fassi, SVP Transformation & Automation, Media and Retail Measurement, and Oleg Royz, Vice President – Global Digital Transformation / RPA. What follows is mainly in their own words, and we provide, at the end, an analysis of Nielsen’s progress.
Adopting RPA and Cognitive Automation
As a very large and dynamic organisation with multiple business units, products and operational needs, Nielsen began looking at new technologies, including AI, particularly machine learning, and how they could advance the business. The question was: “Can they use machine learning?” “Not quickly” was the answer. Which is why Nielsen started to look at other technologies, including RPA. A lot of Nielsen’s manual processes were very repetitive and involved connecting different systems together. It was worth looking more closely at RPA technologies. In the six month search process the most important selection criteria was that the vendor’s solution be able to connect to all Nielsen’s platforms easily and reliably. On the tests for business functionality the vendor UiPath performed better than the competition for seamlessly connecting with Nielsen’s proprietary platform systems; integrating with different tools; and automating business processes. The second criterion was a vendor with a deep commitment to, and investment in, transparency and community. UiPath was differentiated for their strong culture of openness, the UiPath Community of Developers, and the free UiPath Academy with a robust curriculum. Additionally, the UiPath solution came with a customer success manager located in the right geography.
Nielsen received their first robot licenses in July 2017. They built awareness by talking to the leaders of business units and back office (particularly the finance and HR functions). The objective was to find out what the various work streams looked like, how work was being performed and where the pain points were. It was a time of evangelisation—travelling and talking about what RPA is, and showing demos of the automation in action. Good feedback came from business unit leaders and their organisations—they understood the importance of RPA and why it could make a powerful and positive impact. From there, Nielsen created a strong pipeline of ideas and built a community of users we called ‘RPA champions’.
The strategy was always to make this technology available across the entire company. Nielsen asked for ‘RPA Champion’ nominees, and had a lot of volunteers who were given rotational assignments. The lesson? When people are given the opportunity to learn a new skill, get trained, and receive a certification—something that can go on a resume—the result is a lot of excitement and enthusiasm. By 2020 Nielsen had 200+ ‘RPA champions’ spread across 40 countries. Based on these enthusiastic RPA champions and successful group automation experiments, Nielsen were confident that the first Nielsen automations could be delivered by the end of December 2017 and a fast ramp-up would occur in January 2018. Unfortunately, things did not happen as expected, and it took an additional six months to really take off.
Misplaced confidence in quick success was brought about by two errors of judgement:
1. Underestimating the role that business owner influence plays in RPA adoption:
The first real project was the Proof of Concept and it failed. The automation opportunity was huge—to bring RPA efficiencies to our support desk, covering more than 100 countries, with requests ranging from password resets to more complex problems. The project was aimed at automating this work by developing an RPA solution that integrates intelligent classification that would automatically route tickets for faster response times and full automation of simple requests. Nielsen quickly learned that RPA is unlike traditional technologies, which are created with new features and capabilities. RPA, on the other hand, creates powerful solutions by truly understanding the process and using that knowledge to build automations driven by good decisions, whether role-based or by the intelligent calibration of machine learning models. In either case, the RPA team needs buy-in and collaboration from the process owners. As the Nielsen team began to dive deeper into their process automation work, there was cultural pushback. Business owners weren’t ready for RPA—or thought there were better solutions coming from their own organisations. The learning? For RPA to be successful, Nielsen had to have engaged business owners who cared about the outcome.
2. Underestimating RPA skillsets and time requirements:
With a primary objective to democratise RPA, Nielsen thought non-developers could easily learn how to set up and deploy an automation by themselves, and a decentralised model seemed the right approach, with a fully decentralised team of RPA developers from the business, and a light central organisation to manage governance and best practices. RPA champions, identified by business leaders, would build out this model by learning RPA and be the decentralised resources for process automation. But RPA proved not that simple. In fact, most processes require complex setup, with a steep learning curve for RPA champions who are not developers, and—more importantly—are not dedicated to the project full-time. So, despite widespread enthusiasm among the large community of RPA champions and a pipeline full of automation opportunities, Nielsen’s de-centralised model resulted in slow, limited progress.
The silver lining of doing it ‘the wrong way’ was the huge network of people trained in the vendor’s RPA technology. Nielsen now has (in 2020) over a hundred people, in every function and almost every country—all who have been trained by UiPath Academy. They are the RPA ambassadors and identify automation opportunities across the organisation.
Lack of progress led to a move from a de-centralised to a hybrid engagement model, with a central hub of experts and a team of RPA developers and others with the expertise needed to handle engagement security and infrastructure—centralising the decision-making, so teams were no longer tackling those areas by themselves.
The hybrid model places these resources in an RPA Centre of Excellence (CoE), where their expertise has two key areas of additional accountability. One is to coach and mentor the RPA Champion community, enabling champions to own the setup, development and deployment of small processes. The second is to own the development of complex processes (defined as consuming 2,000 annual hours of manual work)—in collaboration with RPA Champions and process subject matter experts (SMEs) helping with process documentation, confirming the ‘to-be’ state, and testing. The collaborative nature of this hybrid model gives champions the opportunity to shadow RPA experts and learn best development practices.
This change in the approach triggered an immediate increase of the throughput of automations. The 150+ RPA champions became the driving force in identifying new opportunities. They have all been trained on RPA basics and work very well with the RPA developers from the CoE.
Business Value Delivered
Once the hybrid model replaced the original de-centralised approach in March of 2018, Nielsen saw impressive progress. More teams across the globe adopted RPA, confidence in RPA grew, and the size and complexity of the processes automated increased. Specifically, in twelve months of the hybrid model ending in March 2019:
- • A team of 8 RPA developers were recruited.
- • Certified RPA Champions grew from 71 to 150+.
- • 70 processes had been automated.
- • ‘Man’ hours automated climbed from 6,000 to 347,000 (over 400,000 by 2019).
Nielsen moved the conversation away from traditional cost and headcount towards hours saved by automation. The result? Reporting RPA outcomes to management became expressed in hours of work automated, and senior management stopped asking for reports on dollar targets.
Business units are shown actual outcomes from earlier projects, and there seems to be an exponential effect—“like a light bulb coming on”, said one Nielsen executive. All this created huge motivation because people started to realise that they would get better quality, better service level agreements (SLAs) and a better customer experience and commitment. Business operatives were able to connect the dots and see how RPA technology could actually be applied in their own world. That change in dynamic and surge in motivation was a huge milestone for the overall automation program. And it all began by talking about hours, starting small, and showing real proof.
The Cognitive Future
Initially the RPA team included a scientist working on machine learning. When automation encompasses the use of AI technology, especially image recognition or a type of classification algorithm, it’s apparent that the lines between AI on one side, and RPA on the other side, become blurred. At Nielsen, the intersection of AI and RPA is where they find opportunities to leverage AI technologies such as image recognition and computer vision definitely, as well as some use of machine learning. For example, intelligent ticket routing. Ultimately, Nielsen’s automation is turning into a continuum of technologies. It’s only very recently that the journey to what’s termed intelligent automation with humans in the loop began. That step brings RPA closer to business process automation integrated with business process management. Those seem to be the logical next steps. Nielsen started with repetitive processes, the routine ones that don’t require intelligence but by 2020 they were leveraging AI technologies to continue making automation impacts.
To unleash the full potential of robotics, the RPA program scaled up during 2019 with natural language processing, optical character recognition (OCR) and with further automation technologies in production and testing. One example was a virtual assistant with natural language programming (NLP). This capability enables the human workforce to interact directly with the digital workforce. The main objectives were:
- • Important way to ‘humanise’ the digital workforce, by enabling a conversation between human and digital colleagues.
- • Reduce the work load on the RPA Support team
- • Allow faster response to questions about the status of transactions executed by the Digital Workforce.
Another use case was visual processing and intelligent extraction of data from documents. In this context, Nielsen implemented the extension of RPA to incorporate capability to extract data from documents and furthermore process it in a touchless manner. In daily activities, documents tend to differ (each customer is using its own format). Besides the generic data, as well specific data need to be identified. The step to increase the quality of further execution is the removal of mistakes. Thus, additional verification had to be incorporated, to make sure that failures would be detected and corrected and not propagated along the chain.
What has changed? Thanks to RPA and respective adjacent technologies Nielsen can move the conversation out of the technical domain into the process domain, focusing on the business challenge and opportunities. RPA allowed us to keep conversations about automation mainly at a process level.
For Nielsen, RPA so far has not been the silver bullet for managing cognitive. The main benefits from managing cognitive automation has come from other software development programs, and ultimately coding. Nielsen sees the links between RPA and other technologies becoming even more important to master, for example using Agile Development Frameworks, as they expand their cognitive capabilities further. Some areas are very well covered on the market, like ChatBot (using Natural Language Processing), some types of Computer Vision (extraction of data out of form documents) and can be almost delivered as plug and play. Nielsen has accelerated expansion into those cognitive areas, because a lot of products and solutions already exist. Cognitive opportunities beyond this will need to involve Data Scientists (to tame the data), Automation Engineers (to tame the flow) and Subject Matter experts (to tame the processes).
How does Nielsen envision the future of work and work to connect the dots? Nielsen have visualised the strategic connect towards ‘Intelligent digital processes’, bringing together people, data, analytics, processes and automation. The strategy when implemented will strengthen the four key areas—augmented workforce intelligence; scaling digital value chain; active business process monitoring; and digital workforce management.
By early 2020, before COVID-19 struck, Nielsen had delivered 20-fold processed transactions growth year on year. To continue growth rates beyond 100 percent will require moving from automations to a digital workforce that need to be managed in a dedicated way. RPA started by providing sub-parts of processes with reduced interactions. RPA is moving to be connected into Business Process Management tools, as an option for integration. Overall Nielsen think they should be moving to the ‘enlightenment’ phase where the technology potential and usage is broadly understood, and productivity is harvested on a much larger scale. RPA will move into becoming a mainstream tool in the company’s automation toolbox.
Looking a few years ahead—beyond the 2020 disruption—more seems to be possible, including self-designed bots, intelligent bots and a true hybrid workforce. Technology to capture processes and build automatically process bots will be available, allowing a much faster development of basic RPA. Today most of the RPA development is based on human coding (a team of programmers need to convert/code from a business process specification). Nielsen have been seeing automated business process capture, incorporating cognitive capabilities, able to document processes, RPA-ready, in the range of 20 percent–50 percent. As these rates increase, ‘code scaffolding’ will be automatically generated, increasing significantly the productivity of bot creation. New bot generations will follow, including Artificial Intelligence capabilities, having moved from rules based into algorithms and self-learning.
In the Nielsen view, the hybrid workforce will provide augmentation for each member of the human workforce. Access will be through a personal Digital Workforce assistant. This assistant will be able to help, with a unified user experience, interacting with natural language (written or talked), in different knowledge areas in a bidirectional conversation. Personal Digital assistants will coordinate the work with a pool of specialised bots, though presenting one personalised interface to the human. This will provide a spontaneous update of the activities the Digital assistant has performed on behalf of the person, identifying which ones would need intervention, and providing adequate contextual information to allow faster correction.
Lessons Learned According To Nielsen Executives
Lesson 1: Recognise at the outset that a hybrid model is an excellent way to achieve good RPA solutions for fragmented processes—and almost every company has a wealth of fragmented processes (as a result of past acquisition or co-existence of legacy systems and new systems). The hybrid model enables companies to scale RPA effectively and achieve the large volume of automations required to turn fragmented processes into a significant impact. Typically, the very first round of process identification yields only small opportunities. Without a hybrid model, it’s a challenge to propose RPA as a viable automation solution to business leaders who have large productivity targets to deliver. In fact, Nielsen are confident that RPA can deliver faster on automation opportunities than any other competing technology.
Lesson 2: Involve IT. Much has been written about the importance of involving the IT department at each step of the process—from the selection of the vendor to the definition of the governance process of the robots in production. For Nielsen, even though they involved security and IT representation in the vendor selection process, they still got stuck with their first robot user ID request. Why? Because large companies suffer from inflexible processes. In Nielsen’s case, getting the userID for the robot took several months and a three-day in-person workshop covering the infrastructure and security aspects. The participation of the vendor at the workshop was a key step in building trust and removing the final concerns of our IT teams.
It’s important to realise that RPA is a technology solution, and as such, it should be set up with the same rigour as any other IT solution. Additionally, the robots are going to interact with platforms in different ways than humans. Even if they perform the same clicks, the speed at which they will do those clicks will generate additional stress on existing platforms and could break them. To avoid disasters, Nielsen implemented a strong governance process with formal approval required from the main platform stakeholders before moving the process into production.
Lesson 3: Engage process owners and subject matter experts (SMEs) early and often. With any digital transformations, and RPA in particular, the key to a good outcome is to truly and fully engage both process owners and SMEs—defining the rules and the framework for their processes. This work can easily be the longest and most challenging phase of any RPA project you embark upon. That’s because, if the data isn’t standardised, or if the process isn’t standardised, the outcome of your RPA use case may be seen as a failure, even though the fault was in the data or activity exceptions. It’s absolutely essential to have a firm commitment for engagement from the people who actually own and understand the process.
Nielsen is a very interesting case of learning from planning and doing. They have made substantial progress on service automation over three years, which has placed them well for accelerating automation through and after the 2020–2021 crisis. This puts them amongst the small group of digital leaders moving into 2021. Focusing on their RPA approach, for us, four insights emerge:
1. The first is that Nielsen understood the value of cognitive technologies but recognised that, in late 2016, the time was more right for RPA. This fits with our own timeline, that even by 2019, there was still massive potential for RPA and that cognitive automation was still in embryonic, proof-of-concept, pilot stage—though Nielsen itself has been working on the RPA/cognitive/AI interface for some time.
2. The second insight is into Nielsen’s strategic behaviour; recognising from the start the need for a solution that integrated with various technical platforms and tools while exhibiting strong business functionality. On selection criteria it is interesting how the Uipath Academy and training issues figured so early in Nielsen’s thinking, indicating their awareness of change management and capability issues, as automation scaled. Even so, they self admittedly underestimated skills requirements—a cautionary note for those no as progressed as Nielsen.
3. Thirdly, on early stakeholder buy-in and governance, Nielsen learned the hard way—by not involving business owners enough, and through adopting an unrealistic governance model. But Nielsen corrected quickly, moving to a hybrid governance model, and building a strong RPA community. This led to impressive multiple business benefits being produced, akin to what we found possible in earlier cases as ‘the triple win’ for shareholders, customers and employees .
4. Fouthly, Nielsen’s top three lessons chime with our findings throughout our recent book, Becoming Strategic With Robotic Process Automation, —get the governance model right, engage the stakeholders early and continuously, and particularly the IT function and its myriad of technical specialists.
(i) See Lacity, M. and Willcocks, L. (2017) Robotic Process Automation and Risk Mitigation: The Definitive Guide, SB Publishing, Stratford-upon-Avon, UK. Especially Chapter 2 – ‘The RPA Art of The Possible
(ii)See Willcocks, Hindle and Lacity (2020) Becoming Strategic With Robotic Process Automation. available from firstname.lastname@example.org