Because Walmart is an industry behemoth, creating a more agile technology stack capable of capitalizing on bleeding edge technologies is a feat.
Key to transformation is honing internal technical capabilities and capitalizing on data, which creates a strong foundation for innovation is central to maintaining a loyal base and attracting new customers.
The biggest retailer in the world — and No. 24 on Forbe's 2018 list of the world's largest companies — Walmart has a vast technology ecosystem spanning more than 5,300 stores in the U.S., featuring robust back office and customer-facing capabilities.
Walmart's 2.2 million associates generate terabytes of data.
That environment lends itself well to artificial intelligence, which requires good, clean data, according to Yazdi Bagli, SVP of global business services and emerging technology at Walmart, speaking in an interview with CIO Dive. If the company can derive insight and apply it to process, it can save money and reinvest in the consumer experience.
All told, "Walmart is like a data scientist's dream," he said.
Putting AI in the back office
Bagli spent 23 years working in shared services at Procter & Gamble before joining Walmart earlier this year. While P&G is a global brand with household names like Pampers and Tide, its scale pales in comparison to Walmart.
P&G began implementing robotic process automation technology (RPA) about four years ago, which automated what a human could do across different platforms. In particular, Bagli oversaw procure-to-pay, which handled 5.5 million invoices.
But Walmart handles more than 200 million invoices. Such a complex environment is ripe for automation and AI.
More companies are investing in AI and the technology is proving to have a strong ROI. More than 80% of executives gained a financial return from AI investments, according to a Deloitte survey of 1,100 IT and line of business executives. Across sectors, enhancing current products, optimizing internal operations and making better decisions are the leading benefits for adopting AI.
Many companies are struggling to apply AI, either not understanding where to start or hoping to scale AI from the outset. Organizations, however, are finding more success in starting small with AI implementations, focusing on main pain points or projects, and scaling over time.
Walmart crafted a compelling business need for AI — taking costs out of back-office systems to invest in the front end — giving the organization a scope for implementation. To start, Bagli wanted to show how AI could impact commodity processes already in the back office.
Commodity processes generally include accounts payable, accounts receivable general ledger and a bit of compensation and benefits.
Data scientists can unleash supervised machine learning on commodity processes. The company can apply data science to ensure that an invoice is 100% accurate, paying the correct amount to the right invoice.
Processing the exceptions
Automation is also a fit for the shared services organizations, which spends most of its time handling exceptions. Anything an ERP cannot do itself comes out as an exception, which is where people come in.
Companies can apply RPA to automate exception handling, fixing problems that are bound to occur before they happen. Bagli prefers to apply RPA upstream, which removes the need for exception handling altogether.
RPA typically automates repetitive, recurring processes, which are routine and logical, according to Mark Davison, partner and global lead, robotic process and cognitive automation at ISG.
When it comes to ERPs, the most automated task involves invoice processing exceptions, Davison said, in an interview with CIO Dive. When an invoice comes in, it is matched against a purchase order and in some cases a receiving document as well. ERPs will perform the match when an invoice comes in, but sometimes that match creates errors.
Business rules to remedy those errors are the same, over and over again, according to Davison. Rather than task humans to follow the business rules, companies like Walmart can program robots to do invoice reconciliation.
Successful technology implementation does not happen overnight. Bagli thinks in two year arcs, and in the next two years he hopes to make at least a couple of big wins automating end-to-end processes.
Long term, Balgi hopes to scale AI across the rest of the company. For example, there is a lot to automate in merchandising operations, which coordinates supplier-retailer interactions and effects things like in-store displays.
"Using AI to really make sure that there is less value leakage" and requiring fewer employees to get the work done is a "huge opportunity in Walmart," Bagli said.
Correction: In a previous version of this article, Mark Davison was misidentified as Davidson.