These types of centers are timely. Automation technologies are the fastest-growing type of AI, even if they are not the smartest. These technologies offer rapid implementation, ease of deployment, and a high return on investment. They are increasingly being combined with machine learning and natural language processing for more intelligence and easier human interaction. There has been particular interest in automation technologies during the pandemic: Applications such as back-office administration and IT operations automation have allowed work to continue even when humans have been unable to come into the office. (A global survey by Deloitte found that 68% of business leaders used automation as part of their response to COVID-19.) Organizations are also using automation as an alternative to outsourcing.

For these reasons, automation COEs play a valuable role. They can help organizations maximize the utility of the systems in use and even obtain competitive advantage from them.

Types of Automation

Automation technologies are quite flexible and are being implemented across a wide variety of settings: They can be applied to structured administrative processes and to well-defined processes such as employee onboarding.

One reason for the flexibility is that there are a variety of automation technology types, including these:

• Generalized automation tools, such as robotic process automation (RPA) for structured workflows involving multiple information systems. This is probably the most popular automation technology category — with the vendors that provide it growing at a torrid pace.
• Tools that support specific organizational functions, such as marketing (for particular marketing tasks, such as campaign management).
• Tools for understanding text from unstructured data in hard-copy documents, leveraging intelligent document processing technology to learn over time and improve accuracy.
• Process control technologies for largely digital processes, including workflow or rules engines.
• Application interaction tools for technologies such as application programming interfaces, enterprise application integration, and enterprise service buses.
• Process monitoring and organization tools for tasks such as process optimization, process data visualization, and process mining, and for overseeing process data repositories.
• Low-code or no-code automation solutions that put basic computer programming functionality into the hands of nontechnical users.

These different categories are expanding in functionality and overlapping, leading to a degree of automation sprawl. Business users of automation can find it confusing to sort out the different categories and their best application areas. An automation COE can determine the vendors and technologies it wants to endorse in each category and help business leaders determine which of the categories makes most sense for their application.

COEs Develop Strategies for Automation

If strategy efforts are to be successful, automation COEs need, first and foremost, to be established with an overall vision and direction for the program, defined by leadership and embraced by the business. With clear direction and guidance aligned to driving business goals and objectives holistically, scaled automation programs have proved time and again to drive impressive results and measurable financial returns.

One key role that an automation COE can perform is to strategize how to use automation within the business in ways that ensure a return on investment. The automation COE for New York-based retirement financial services company Voya Financial, for example, uses a three-stage approach, including discovery (“Should we use automation at all, and should we improve the process first?”), implementation, and assessment of how the process is working. It has a combined methodology for Lean Six Sigma process improvement and automation of processes, pairing automation with a broad approach to process improvement and management. That combination helps to ensure that bad processes won’t be automated and that automation will at least lead to operational performance improvements.

Another strategy is for the COE to focus on scaling automation as rapidly as possible. That’s the tactic taken by Nielsen, the U.S. company that tracks what people listen to and watch. Its RPA center of excellence wanted to have a swift impact on the business. The center supplied a dedicated team of RPA developers — picking up the cost so that managers throughout the company didn’t have to dedicate their own resources — and focused on time savings rather than on cost reduction. In just three years, the company has saved business functions and units about 400,000 hours in manual labor.

Other strategies for automation do involve cost reduction. That’s what New York-Presbyterian Hospital has done, using RPA and machine learning to reduce administrative costs. The focus thus far has been on financial processes, including revenue cycle improvements, ensuring correct coding of patient treatments, and processing denials of insurance claims.

Relatively few organizations thus far have employed an explicit personnel reduction strategy with automation, although in a 2018 Deloitte survey on AI, 63% of U.S. executive respondents indicated that they wanted to automate as many jobs as possible to reduce costs. But most organizations we’ve encountered have said that their automation is generally focused on freeing human workers to do more creative and interesting work.

COEs must continuously manage their robotic digital workers after deciding which strategy to pursue and deploying them to production. Some believe that once an automation system is built, the work is done, and it will run autonomously with no oversight. This isn’t accurate. A well-run digital workforce maintenance program should not only ensure that the automation systems are optimized to deliver at their peak performance but should also monitor and report on the measurable value they deliver and their holistic impact on business processes.

Other Attributes of Successful COEs

Automation COEs perform a variety of other functions in addition to driving strategies. They may develop automation projects themselves (as Voya and Nielsen do) or train users to do their own development. Dentsu, a large Japanese ad agency, is pursuing a scaling strategy involving educating a large group of “citizen developers” who will create small automation use cases on their own. Dentsu has also appointed a chief automation officer to oversee both citizen and professional development through its COE.

Regardless of where the development activity takes place, a COE should attempt to aggregate the technology’s benefits and value to the organization. Whether the focus is on saving costs, eliminating low-value human labor, or reducing errors, the measurement and aggregation of processes are necessary to establish whether automation is providing value to the organization. If the measures suggest that value isn’t being achieved, the company can take a different approach to implementation.

Automation COEs should also collaborate with business leaders on understanding the implications of automation on the workforce and perhaps coordinate with the human resources organization on employee retraining programs. That type of collaboration can help to determine what people should do with the time freed up by automation, under what circumstances automation leads to personnel reductions, and what types of retraining and reskilling are necessary for existing workers.

Most organizations will have more effective automation programs if they define a philosophy of what humans do best and what machines do best, and how the two will coexist. In general, a “humans with machines” work environment will engender greater levels of collaboration and engagement from employees.

Where Should a COE Be Located?

One last issue is where to place an automation COE within the organization. We see three primary alternatives, each of which has been effective in companies we’ve observed.

One is within the IT function. Procter & Gamble, for example, has a smart-automation COE within its IT function. It works closely with business units and functions to get sponsorship and buy-in for new automation processes.

Some companies put automation capabilities and management within an artificial intelligence or data science group. That helps to facilitate the development of intelligent automation capabilities combining, for example, RPA with machine learning. A North American bank, for example, locates its extensive automation projects within an AI COE. Its projects increasingly combine RPA, machine learning, and other forms of AI. This kind of comprehensive intelligent automation COE allows an organization to reimagine business processes and apply a set of integrated AI and automation tools to them — sometimes referred to as a hyperautomation approach.

Locating an automation COE in a quality, process improvement, or global business services organization is a third possibility. This location is perhaps the best option to facilitate process improvement before and after automation. Business processes are the best way for an organization to plan and manage workflows, and we believe that intelligent automation technologies could lead to a resurgence of process thinking, improvement, and reengineering in businesses.

Wherever a COE is located organizationally, automation technologies will only grow more common and more intelligent in the future. They are an important tool to augment the work of human beings and can free us up to do more imaginative tasks. The technologies, and a center to advocate for and support them, deserve a place in every large company.

This article was originally published by MIT Sloan Management Review on May 26, 2021.