How Vale Struck Gold with AI-Powered Automation

Automation Anywhere delivers a plethora of AI-powered automation innovations today to help organizations automate more processes by moving from discovery to automation in record time. But, every successful automation begins with first understanding the nuances of the process to be automated so you can find the fastest path to process optimization and business ROI.

Vale, a global mining company based in Brazil, used Automation Anywhere's Process Discovery to better understand acquisition processes, identify ideal candidates for automation, and automate more than 3,000 processes to save over 121,000 hours annually. And, it has its sights set on using automation to help workers free up time to focus on more rewarding work. Read on to learn how.

Doubling volume without increasing workloads

Vale mines iron ore, pellets, nickel, and other resources, runs operations to generate copper, gold, silver, and other materials, and has operations in logistics, energy, and steelmaking. It also acquires iron ore so other mining companies can take advantage of Vale’s vast global logistics networks, including facilities and offices in Canada, Indonesia, Malaysia, China, the U.K., and beyond.

Vale is a decidedly massive organization, with over 120,000 employees and tens of thousands of additional third-party workers involved in operations. Of those, 80,000 workers use computers for processes that are potential automation candidates. Finding even small opportunities for automation would easily save the company thousands of hours and millions of dollars.

Couple its size with Brazil’s notoriously bureaucratic regulatory requirements and Vale, which began its automation journey in 2018, didn’t have to look far for automation impact.

“There are several producers in Brazil that sell to us in order to export their products to the world,” Alexandre Pigatti, Head of Artificial Intelligence and Data Democratization at Vale, explained at Imagine Austin 2024. “In Brazil, we have much documentation involved in these transactions. We have plans to double the number of those transactions, but there is no room to double the number of people required to do it. That was our challenge.”

Understanding how things actually get done

At first, Vale used traditional interview-based process discovery methods to define and analyze its existing processes. It was a slow, laborious process that required much rework and follow-up interviews.

“It was tough to schedule meetings because everyone has their own work to do, and they’re concerned with due dates,” Pigatti recalled. “They’re also human beings and they forget things, so we would eventually find gaps when we tried to implement. That was bad because we would need to interview them again. It typically took about three months for this phase.”

Vale saw an opportunity to modernize this critical precursor to automation with Automation Anywhere’s AI-powered Process Discovery. It works by deploying a lightweight software agent onto a user’s desktop to record every process touchpoint. Process Discovery identifies, structures, and maps the process data to create process blueprints and process definition documents (PDD). Automation teams can then analyze the processes to find optimization opportunities, decide on next steps, and inform compliance efforts.

For this regulatory documentation process, Vale installed Process Discovery agents on the computers of just seven workers and let it run for 10 days. In that time, it captured 370,000 events.

“With Process Discovery, it took just 10 days, and that was much better for us,” added Pigatti. “And Process Discovery was better because it never forgets a step.”

From three months to 10 days, Automation Anywhere helped Vale reduce its process discovery cycles by 89%! But that was just the first of many savings.

Optimizing and automating the entire process, even steps outside the norm

Process Discovery also produced valuable process insights Vale may have otherwise missed. First, it found little process variation across the users, indicating the process was very stable, mature, and ideal for automation. It also found opportunities for optimization in a step that required the sellers to complete a quality control form. Vale expected standardized answers, but each seller filled out the form in a different manner, indicating an opportunity for additional training or to improve the form itself.

At the system interaction level, Vale expected workers to complete most of the process in its SAP system. Analytics from Process Discovery actually showed that workers used Microsoft Excel, File Explorer, and Outlook more than SAP, signaling additional optimization opportunities.

“Workers used Excel 42% of the time, so we wondered if we should better train workers with SAP or make Excel part of the process,” Pigatti said. “It was an interesting finding because, had we just done process mining, it only works on logs from SAP and other legacy systems. We don’t have logs for Excel or Microsoft. Process Discovery captured everything.”

Exploring new ways to transform work with automation

With these results from Process Discovery, Vale identified six process categories, from order creation and invoice management to product quality reporting and fulfillment.

“Automation wasn’t about reducing workforce; it was about freeing time for people to do more analytical and strategic work,” Pigatti explained.

“We were able to save 2,500 hours per year—that’s $5 million in cost avoidance without increasing FTEs—and it totally changed how the entire department worked.”

Vale is currently midway through automating its materials acquisition process, yet has automated over 3,000 tasks to save 121,000 hours annually across more than 30 areas. And there’s more to come. With 810 citizen developers trained, Pigatti is looking towards additional cost and time savings and opportunities to automate more processes across the entire organization. A big component of those efforts will be Automation Anywhere Automation Co-Pilot, the first generative AI-powered assistant engineered for automation.

“We’re looking to apply Co-Pilot across several legacy systems,” said Pigatti. “We want to put generative AI into action. And, we have retrieval-augmented generation (RAG) services ready to be put into action as well.”

Pigatti mentioned examples where Vale plans to use AI-powered automations and chatbots to perform tasks more quickly and accurately.

“Imagine asking a chatbot to schedule my vacation. The chatbot could go to the HR system and schedule my vacation (because I always forget how to get into that part of the system). The chatbot would then alert me when the process is completed. Or, I might want to know the production from a specific iron ore mine. I could ask the chatbot, and it could query our manufacturing systems, generate the reports, extract the information, and answer me.”

“It’s these types of automations that will change the way our people interact with our systems,” Pigatti concluded. “It will completely change how we work.”

To learn more about Vale’s AI-powered automation journey, watch the Imagine 2024 session, “Vale Demonstrates the Power of Process Discovery Driving Greater Efficiency and Precision”.