Harnessing Grassroots Automation

Companies are increasingly embracing the idea of helping nontechnical staff members — those who have deep business-area expertise — learn to directly automate processes that give them headaches and eat up their time. For instance, human resources employees are uniquely qualified to identify the mundane and repetitive parts of their jobs, such as candidate-tracking tasks, and then, with some training, build automations that will relieve them of chores such as duplicative data entry and data cleaning.

While the development of such applications by so-called citizens within organizations requires careful planning and governance to be effective, low-code and no-code technologies have become commonplace and made such ventures possible.1 Specifically, robotic process automation (RPA) and a broader intelligent automation (IA) suite that allows for the redesign and automation of workflows are now straightforward enough that functional experts can design, develop, and deploy IT applications and analytical models themselves. No longer do all projects require mediation by IT employees, who might not fully understand end users’ pain points. These tools of citizen-led automation are allowing less-technical people to build complex systems that improve their work experience, and they are already generating considerable value for many businesses.

In this article, we draw on interviews with six companies — AT&T, Dentsu, Johnson & Johnson (J&J), PwC, Voya Financial, and Wesco — to describe their efforts to join the citizen automation movement. We also detail how other organizations can best develop these capabilities and the benefits and challenges of doing so.

What’s Driving Citizen Automation

At its core, encouraging non-IT professionals to participate in designing their own work tools is not new. Enterprises have long tapped into teams across their businesses for process improvement ideas. Six Sigma belt wearers, for instance, have been trained in improving small processes. What’s new is that today’s citizens can actually sketch out and then run the future state they were once only able to describe to IT development teams.

Harnessing citizenry is partly necessary because there are simply not enough people with the professional IT skills needed to accomplish the torrent of digital initiatives on companies’ agendas. Even conservative estimates project a dramatic shortage of tech workers by 2030.2 Tasks such as moving information between transactional systems, updating spreadsheets, and even composing standard-format emails are ripe for automation, but that often doesn’t happen because people with the skills to do that work aren’t available.

At accounting consultancy PwC, the citizen automation effort arose out of an initiative to train employees who would be known as “digital accelerators” to help grow the business without adding a proportional number of staff members.3 Data, automation, and AI were identified as the three pillars of the initiative. Employees who volunteered to learn new skills and technologies, and were selected for the program were asked to focus on one of the three pillars. Participants were given time off from their regular client service jobs to pursue a variety of upskilling options. Employees who focused on automation were trained in the tools (including RPA, data prep and blending, and simple machine learning models) and in Six Sigma process improvement methods and were asked to identify processes that would benefit from automation. One audit-focused digital accelerator created an automation to scope needed tasks on a client audit by automatically extracting and aggregating data from many different spreadsheets. It saved 40 hours from the audit engagement and was adopted as a standard tool for auditors to employ with other clients.

Of course, a primary driver of the rise in citizen-led automation is the corresponding dramatic rise in the relative simplicity of some automation programming.

Necessary Tools and Training

The citizen automation movement is enabled by the rapid evolution and democratization of automation tools. Compared with other forms of artificial intelligence, RPA and IA tend to be easier to implement and less expensive. RPA is being widely adopted to access data from multiple systems and to automate structured, information-intensive tasks, such as routing incoming customer emails or updating order status in a transaction system. When combined with IA tools such as machine learning and character recognition, they can also make data-driven decisions and extract important information from documents such as handwritten customer forms or key provisions in a contract.

There are several technology options. First are standard RPA tools from vendors such as UiPath, Blue Prism, and Automation Anywhere. These can be complex to learn and use — not because they require coding but because they may need to be integrated with transaction systems. With proper training, many nontech employees can build simple automations with them. Second is technology that is specifically developed for citizens and thus involves little or no coding. Some mainstream RPA software comes in simpler versions intended for citizen automation. Microsoft, for instance, has made easy-to-use RPA capabilities available as part of its Office productivity suite.

Most of the organizations we’ve spoken with in our research offer citizen automation training programs that are between 40 and 80 hours long, and many supplement their programs with instruction provided by the leading automation tool vendors. Training programs can also include hackathons in which trainees apply their skills to quickly build RPA applications.

Because RPA systems typically link to and extract data from existing transactional systems, citizen automators often need to have an awareness of corporate IT architectures in order to safely access and use data. However, if citizen-developed RPA applications are certified by IT professionals and address any issues around integration with other systems, this knowledge might not be necessary. Some organizations we’ve studied have established automation centers of excellence (COEs) that handle all such integrations and compliance, allowing citizens to focus on applying automation to the processes they understand without requiring them to become familiar with the complexities of the underlying architectures.4

The recent appearance of generative AI on the enterprise scene is already beginning to make RPA design and implementation easier. Since OpenAI’s ChatGPT was announced in late 2022, for example, several RPA vendors have announced interfaces between their RPA systems and the language capabilities ChatGPT offers. Before long, it should be very easy for a user to specify the desired attributes of the automation system in virtually any natural language and have a working prototype of the system automatically produced. The generative AI system should also be able to automatically create an easily understood description of the workflow and decision rules, if prompted to do so.

Companies pursuing intelligent automation may make machine learning capabilities available so that they can be combined with RPA. In most cases, this will involve automated machine learning, or AutoML systems.5 These systems automatically perform many of the tasks involved in data science, including minor data cleaning, feature engineering, alternative algorithm modeling, and code/API generation for deployment. Some AutoML systems are technically complex, but others are more suited for use by citizen data scientists who have only some quantitative training or background knowledge. At AT&T, for example, citizen data scientists can work with predefined customer attrition data and features to identify which mobile phone customers should be targeted with a promotion or some other intervention to prevent churn. An RPA bot can then automatically notify those customers about the promotion.

It is almost always a good idea to improve processes before automating them, and training citizens in incremental process improvement techniques such as Six Sigma or Lean can be beneficial.6 A centralized group of process improvement specialists can also provide a quick process analysis before the automation is rolled out. However, there is a case to be made that automating existing “messy” processes still provides value. It might also be the most likely scenario in early citizen-owned efforts, given that they are prone to focusing on digitizing what they do and how they do it — albeit with new tools and technology, in this case.

There are a variety of approaches that companies can take to selecting employees to participate in automation activities. Some, like electrical and communications technology distributor Wesco, happily enroll all interested citizen automators in training. The leader of Wesco’s program told us that the company is willing to train anyone who expresses interest and that he believes the organization even benefits from program dropouts, who can still act as “automation ambassadors.”

Other companies, like PwC, have a formal application process for evaluating and “admitting” volunteers. The pharmaceutical and consumer goods company J&J requires that citizens working with RPA complete a formal training curriculum and be certified in the technology. Leaders there say that, in general, the company sees people with a logical mindset, technical competency, and an aptitude to learn as good candidates for becoming citizen developers. J&J also seeks a fit relative to the work that the employee is involved in, with rules-based work being particularly desirable.

How Leading Organizations Support Citizen Automation

Some companies support citizen automation from the very top of the enterprise and coordinate the activity centrally; others are content to let a thousand flowers bloom more autonomously. Most, however, employ some degree of centralized coordination, technology standards, and training.

In order to magnify the impact of citizen automators, several companies have established accessible collections of automation programs developed by employees for use by their peers. PwC has created one of the most extensive libraries, with several thousand automation and digital assets aligned to client service maps. Before an asset is added to the marketplace, the solution and its underlying code are reviewed in terms of its technical capabilities, its potential business value, and its compliance with policy guidelines. It is then “checked in” to a library for others to withdraw and use.

Global ad agency Dentsu, which was one of the earliest developers of a citizen automation program, established an automation hub where its citizen automators can contribute the solutions they have created. The company now has more than 1,000 actively contributing participants, with 350 solutions in its marketplace. Popular citizen-led submissions in the hub include automations for preparing and manipulating spreadsheet-based reports, exporting and filing invoices, and handling timesheet reporting and email reminders, along with various approaches to extracting data and moving it between systems.

Wesco’s approximately 200 citizens work with the support of a dedicated COE that supplements their automation efforts by handling more technical tasks, reviewing and signing off on automation designs, and ensuring compliance with corporate governance and security protocols. In one program, the COE built a framework for citizen developers to configure the pulling of data from supplier portals and websites to enable better real-time shipment and order-fulfillment tracking. With over 50,000 suppliers, all with unique interfaces, this task would have taken hundreds of IT staff members. Instead, citizens are able to slowly chip away at digitizing the supplier input — a task that is low risk and lower complexity. The COE is then able to build on this by interfacing with the company’s core systems to provide timely information to its customers. As the proj­ect sponsor put it, “The beauty is that everyone is doing what they’re good at.”

At telecommunications company AT&T, both citizen automation and citizen data science are broadly encouraged, and there are active communities in both areas.7 Citizen automation developers who want to engage in more sophisticated machine learning-based decision-making can employ a variety of AutoML tools. AT&T also has a variety of reusable data sets that can be used in machine learning, and a “feature store” with predefined features for machine learning models. For example, users can easily execute searches to find all of the features that involve “churn” or “network performance.” They can also access already computed scores for common predicted outcomes, like churn, via APIs.

The most successful citizen automators are celebrated and financially rewarded at some companies. J&J holds an annual showcase in which citizen developers receive recognition for the impact their solutions have had on business outcomes. At PwC, those who succeed in creating an automation asset for use by the overall organization are rewarded financially if it is widely used. It’s important that employers recognize citizens — financially or otherwise — for the benefits their work has generated, or they may try to take it elsewhere. (For instance, in online discussion boards for individuals who have multiple remote full-time jobs, we’ve come across several participants who have been able to produce sufficient work outputs by automating their own tasks.)

Strategies for Managing Citizen Automation Programs

Effective citizen automation on a broad scale doesn’t just emerge spontaneously. It needs to be facilitated by support functions, encouraged through community development efforts, and guided with policies and guardrails. Citizen efforts should also be monitored and their results tracked in order to understand whether they are valuable and growing. And, perhaps most importantly, citizen automation efforts must be valued and sponsored by the organization’s leaders.

At J&J, to better coordinate automation initiatives, a partnership was formed among the company’s IT organization, its Global Services function, and a few of the citizen developer practitioners. The partners developed a framework to strengthen the support for citizen developers through curated technical training curricula, a set of technology standards, and criteria for assessing use cases.

AT&T has developed an active citizen automation community, supported by a central data science and automation organization. Automation began in a business operations group and rapidly grew to about 300 RPA bots. After a year of bottom-up activity, a corporate COE was formed. Now A&T has more than 3,000 bots, with 92% of the automation use cases built by citizens. The central group develops automation technology standards and holds regular meetings and training sessions for those who wish to join the community.

The PwC digital accelerator initiative was strongly supported by Tim Ryan, U.S. chair and senior partner. Citizen automators in that program have broad discretion about which automation activities they pursue, but only the most successful products are promoted by the firm’s centralized Products & Technology group. That group’s leaders decide which citizen initiatives become enterprise assets for the organization overall.

The Wesco citizen initiative has no single executive sponsor; rather, it is supported by the entire leadership team and coordinated by Maxim Ioffe, the global intelligent automation leader. Ioffe is also responsible for the automation COE, thereby providing a single point of contact for collaboration with IT, advanced analytics, IA, and citizens from among the business teams.

Our research shows that most companies that are taking citizen development seriously invest in the creation of automation COEs that are staffed by professionals with RPA and IA development backgrounds, to accelerate citizen participation. A subset of these businesses, like investment company Voya Financial, have combined automation and process improvement capabilities in one support organization.8 These centralized resources are important for large or strategic business processes that demand deep automation and process improvement skills. They can also offer training (in both citizen-oriented tools and process improvement), certification, and automation project review for citizen activity.

Dentsu started its automation approach with top-down oversight. The initial intent was to automate large-scale enterprise processes, but an analysis found that there weren’t enough of them to make them the primary focus. Instead, the company’s first chief automation officer (one of the first such officers at any company), Max Cheprasov, concluded that there were many more opportunities to automate smaller, local workflows and that the work could largely be done by citizens. Cheprasov’s group developed training, conducted hackathons, and developed a centralized hub for citizen automation projects. He was so convinced of the efficacy of the bottom-up approach that he brought it over to the midsize marketing services company where he now works.

Weighing the Risks and Rewards

There are risks to citizen development. Citizens may automate bad process workflows, fail to document their efforts and leave the company, make mistakes in developing automations, or integrate their bots poorly with enterprise transaction systems. Of course, professional developers can make such errors too, but there are usually other professionals around to back them up.

A key step, then, is to certify citizen-developed automations for ongoing use by the entire department or company in an automation hub or marketplace. That certification can help ensure that the automations were well constructed with state-of-the-art tools and that their function and risks are well understood. PwC, for example, has a governance process that vets citizen automation submissions to ensure that they comply with legal and regulatory requirements. AT&T also has professionals review citizen efforts before they are put into production deployment.

But one leader told us that his governance concerns extend more broadly, into business continuity and change management. “Have we appropriately prepared for 30% of finance being run on citizen-developed applications?” he asked. “We realize a tipping point is coming and wonder if we have the appropriate controls in place for that event. We’re not sure we do.”

While companies hope that citizen-led transformation will create significant value, in most companies the scope of automations to date has been small and focused. Scaling these pilot efforts to address end-to-end enterprise business processes will be uncharted territory. Will citizen automation smoothly enable new efforts in digital transformation, or create a chaotic landscape of shadow IT (applications developed without the knowledge or expertise of the IT function) and technical debt (applications that rely on obsolete technologies)? Even the early adopters we spoke with aren’t sure. At the moment, the movement toward citizen automation seems generally positive, but problematic issues may appear over time and with greater adoption.

These risks are offset by rewards. The most common objective of automation programs, according to a Deloitte Global Intelligent Automation survey, is cost reduction: Companies expected an average of over 30% reductions in costs for automated tasks.9 Companies also highlight the benefits of freeing up people to perform higher-value tasks rather than automating away jobs. Some companies measure employee work hours saved. Dentsu, for example, counted more than 400,000 employee hours saved from its citizen initiatives. AT&T computes that its automations, the great majority of which are created by citizens, save about 270,000 hours per year, yielding a 20-times return on investment overall. In addition to benefiting from these time savings, organizations typically find that employees are more invested in improving and digitally transforming their work.

There are distinct nonmonetary benefits to citizen automation as well. Employees gain new skills and spend less time on drudgery. Task digitization happens faster than it would otherwise, and more employees get engaged in making change happen.

What’s clear is that user-friendly digital tools are driving the creation of a new class of hybrid business/technology developer. These employees are now able to dream up and deploy solutions more quickly and easily than at any time in the history of modern enterprise.