Spotify has found new ways to embrace innovation by deploying more RPA processes using its Center of Excellence, enterprise developers, growing a team of citizen developers, and by identifying new metrics to demonstrate ROI.
Spotify has always been a company whose success has been built on innovation. Since the company launched its global audio streaming and media services in 2008, it has been adapting technology to improve the user experience, from curating playlists to suggesting new podcasts. They apply that same commitment to harnessing technology to simplify business processes as well, including adopting robotic process automation (RPA) to streamline operations.
Spotify first started to experiment with RPA in late 2017 for Treasury processes. With executive sponsorship from Craig Butler, Vice President, Financial Engineering Leader, Spotify initiated a comprehensive RPA selection process, choosing a UiPath competitor. It was during the first year of RPA development that Spotify formed its RPA Center of Excellence (CoE), with a mission to focus on RPA coding, IT governance, development, security, and creating a service center.
After initial success with another RPA platform, Spotify decided they wanted to move away from a code-based solution to a platform that would allow accounting and other departments to develop their own automations. Their automation solution needed to support their vision, and align to the company slogan, "Harder, Better, Faster, Stronger." In April 2019, the CoE hosted a cross-functional workshop with participation by the departments impacted by RPA, including Finance, Customer Service, Legal, Tech, and Procurement. The consensus was that Spotify needed a more scalable platform that was easy to use, flexible, more secure, compliant, and easy to deploy across the entire organization. In Spring 2019, the company conducted a second vendor selection where UiPath was the clear winner, and UiPath was on-boarded in August 2019.
In making the transition to UiPath, improving usability and increasing scalability were the primary goals. They brought in a delivery team of consultants to help with deployment and development. However, the Spotify tech team established goals, determined strategy, and maintained control over the development process.
Sidney Madison Prescott
Coming out of our RPA upskilling program, our people have an automation-first mindset.
Global Intelligence Automation Leader, Spotify
Building a Citizen Developer Program
Once the CoE team had deployed UiPath, the team successfully established the foundational layers for its successful RPA program, including the operating model and citizen developer program, and migrated 11 existing processes to UiPath. The Accounting team became the pilot group of citizen developers and after developing automations for more than two years, the global business unit is looking at ways to scale RPA with unattended bots.
"We really focus on providing automation in two different ways: through enterprise-led automation, which is all of our unattended bots, and citizen-led development, through all of our attended bots," said Sidney Madison Prescott, Spotify's Global Intelligent Automation Leader. "We're basically taking our accountants and upskilling them to build their own automations alongside the more robust, end-to-end automation that our Center of Excellence developers are simultaneously producing. Coming out of our RPA upskilling program, our citizen developers have an automation-first mindset."
"Spotify's decision to expand their program to scale citizen-led automation was an important strategic choice. It allowed them to engage business users in a more hands-on manner and put more implementation wins on the board, which then furthered organizational buy-in behind their enterprise efforts. We see too many clients struggle to gain the right momentum when their output is constrained by their own centralized resources," said Kevin Kroen, Intelligent Automation Leader, PwC U.S.
UiPath played a big role in the success of the burgeoning citizen developer program. UiPath started citizen developer instruction onsite, and when the COVID-19 pandemic hit, the Spotify citizen developers were able to pivot to UiPath Academy and use virtual training sessions.
"Training is absolutely essential, and you need to adopt a robust training framework," Prescott said. "We began that journey by leveraging the UiPath format and that's been very successful for us in avoiding the hiccups we anticipated with upskilling and the rollout of the citizen developer model."
Spotify's decision to expand their program to scale citizen-led automation was an important strategic choice. It allowed them to engage business users in a more hands-on manner and put more implementation wins on the board, which then furthered organizational buy-in behind their enterprise efforts. We see too many clients struggle to gain the right momentum when their output is constrained by their own centralized resources."
Intelligent Automation Leader, PwC U.S.
RPA Growing Pains
Looking back over the RPA work done to date, it's clear that having a robust change management system in place has been crucial to the program's success.
"It's essential that you have a properly defined software development lifecycle which adheres to internal processes, systems, and controls," said James Quinn, Global Intelligent Test Automation Lead, Spotify. "We are subject to stringent financial oversight and control, right the way through our ecosystem, so we need to remain transparent and compliant in everything we do. A robust Change Management process has enabled us to do this."
With the change management and lifecycle protocols in place, the team is focused on increasing the velocity of the development process. Moving forward, the plan is to shorten the development lifecycle by embracing quality initiatives within the CoE, including establishing design and development guidelines, firming up quality assurance and user acceptance testing (UAT), and embracing test automation. Test automation has become a core pillar supporting RPA development, and is being offered as a-service to other teams in need of end-to-end testing across the system and integration landscape at the UI and API test levels.
To effectively run unattended bots, the team had to find a solution to manage robot identity and access management. Like most in the industry, Spotify uses two-factor authentication for all interfaces. For the bots to access the Spotify SaaS systems, they had to use two-factor authentication, so the team tried different approaches. Ultimately, the CoE team started treating the bots as employees, using group structures, with the bots at the lowest level of group membership. That way, the enterprise could use LDAP group permissions to control SaaS application access.
To control automation of SaaS applications, the bots are added to Workday processes just like any other employee for two-factor authentication. Rather than using a smartphone for authentication, the team experimented with IP whitelists to give certain bots access to specific services, but Spotify found this approach wasn't scalable. Instead, they listed the robots with the lowest level of privilege and grant access privileges as needed by using Okta authentication and manipulating data blocks.
Focusing on Company Returns
As the Spotify team continues to expand its RPA program, they also have had to find new ways to measure success. Rather than looking at the number of new automated processes, they are focusing on the business value that automations create.
The Spotify automation teams are assessing the real value of RPA at both the individual and the enterprise levels. ROI isn't just measured in hours, but also in improved accuracy, enhanced employee satisfaction, and increased availability. For instance, unattended bots check exchange rates at off hours and hedge when the rates are favorable. The goal is to determine the true value of automation over the lifetime of a bot or process.
"We have been able to build out some pretty incredible data visualizations and dashboards to show how much time and business value we are giving back to the business," Prescott said. "That's really helping us prove the value of not only the enterprise-led unattended bots but also the citizen developer attended processes."
As the team continues to work with UiPath, they keep sharing insights and learning about new functionality as the platform evolves. The CoE team is incorporating new features, like UiPath Process Mining, Test Suite, and Document Understanding. "I am super excited about adding Process Mining to our tech stack. I think that will become key to help the business identify inefficiencies and process execution gaps, as well as opportunities for process improvements and automations," said Jonas Winqvist, RPA Automation Lead, Spotify.
The CoE team plans to keep improving best practices, creating better code review processes to optimize and speed up overall development. "It is not always about adding more people to support Spotify's continued growth, but also improving how we work - using and tweaking processes and systems effectively to enhance optimization and ultimately velocity," said Quinn. The goal is to continue to anticipate the needs of the company's business users and explore new automation possibilities to meet those needs.
Sidney Madison Prescott
We have been able to build out some pretty incredible data visualizations and dashboards to show how much time and business value we are giving back to the business. That's really helping us prove the value of not only the enterprise-led unattended bots but also the citizen developer attended processes.
Global Intelligence Automation Leader, Spotify
Hear more from Sidney Madison Prescott and Kevin Kroen at our Reboot Work Festival.
Sidney Madison Prescott presented on the stage at FORWARD IV. Please register for 'The Best Bits' to hear more about Spotify's RPA success story.
BY THE NUMBERS
- 100+Bots in production
- 45,500+Hours saved
- 24,000 HoursCapacity released