As I’m saying (and hearing) often at Workday Rising—we see our customers as partners in innovation. They understand their business challenges on a very deep level, and we understand how to apply Workday’s Power of One to help address those challenges.
This is why we embed and apply machine learning (ML) in everything we do, rather than delivering it as a separate application—a strategy that we believe best serves our customers. We know that with Workday as their core system for financials and HR, our algorithms are constantly fed well-structured, single-version enterprise data and associated metadata—the best possible fuel for machine learning. Not only that, each transaction completed in Workday represents a business decision, an outcome that the machine can use to refine and sharpen its predictions for the specific needs of that business.
Our machine learning algorithms learn from this data and customer decisions to offer better predictions. Better predictions lead to better judgement and on to better decisions and business outcomes. It’s a virtuous cycle built on a foundation that standalone ML products just can’t deliver, and is critical to delivering unambiguous and undeniable business value.
So how are our customers using the machine learning integrated throughout all of our infrastructure to make better business decisions? Read on.
Understanding Your Company’s Lifeblood: Finance
When it comes to the financial side of a business, our “lighthouse feature,” as my colleague Matt Grippo puts it, is journal insights. This Workday Financial Management capability automatically surfaces miscodings hidden in the general ledger, dramatically reducing the time and overhead spent by a finance team to close the books.
In use by early adopter customers including fitness and wellness company Life Time, journal insights uses machine learning to detect anomalies in accounting entries by comparing them to other entries for similar transactions. Because we flag these in real time, users can correct potential reconciliation issue.
Katie Van Hauen, senior accounting manager, Life Time, says, “The end result is time back. Through journal insights and machine learning, a process that used to take a day or half a day can now just be a 30-minute refresh of a report so we can know where we need to focus, where we need to put our time and attention.”
Gaurav Sharda, director of Workday operations and process improvement at Life Time, noted: “The more time given back to our business folks—it’s more time given back to our customers. A great customer experience is what we strive for, and Workday helps us achieve that.”
The technology under the hood of journal insights, the foundation on which it rests, is something we call “intelligent core,” an ensemble of machine learning capabilities that leverage state-of-the-art neural networks, and classic machine learning techniques like gradient boosted decision trees to create a powerful predictive representation of the most fundamental element of any financials system: the financial transaction.
With journal insights, we use the intelligent core to find mistakes in the ledger to greatly speed business processes. But that’s just the beginning, because we can also use intelligent core’s understanding of a transaction to supercharge a number of financial flows, including record to report, cash to close, procure to pay, and expense to reimburse. This increase in efficiency and oversight lets finance leaders and their team get out of the weeds and focus on the mission-critical strategic activities that will drive the business forward. For example, journal insights enables a number of automation and reconciliation features that we think our customers will find extremely useful and time-saving, including:
- Supplier invoice automation: Intelligently routes invoices with potential issues to workers who have shown aptitude at resolving similar questions. In its initial release, supplier invoice automation leverages rules-based work queues and header-level scanning to direct invoices to the right person, and can also handle invoices that come in via robotic process automation (RPA) or any other means.
- Expenses via natural workspaces: To remove friction and allow workers to stay in the natural workspace, they can submit expenses using Workday for Slack, and soon, Microsoft Teams. Then, Workday Expenses applies optical character recognition (OCR) to the receipt to fill in data that would otherwise need to be entered manually.
- Customer payment matching: Uses machine learning to process cash receipts from customers that do not include sufficient detail for obvious invoice matching and identifies the likeliest invoices against which to apply the cash receipt.
And soon, customers will be able to benefit from additional automation, including:
- Time series plan predictions: For time series forecasting of plan data to build predictive forecasts that include confidence levels. These predictions can be the starting point for a new plan or could be used for benchmarking or comparison.
- Intelligent expense audit: Will automatically review expense reports and identify unusual activity that requires further review to prevent abuse or fraud.
Skills Are the New Currency in a Changing World of Work
When it comes to Workday Human Capital Management (HCM), we leverage machine learning as part of our underlying platform to drive a wide set of capabilities. As you look closely, you’ll notice that these features all share a common theme: skills. At the very core of our HR investments, we start with our skills cloud.
The skills cloud was built using neural probabilistic language models to map the relationships between more than 200,000 skills. On top of this foundation, we brought additional searching, reporting, measuring, and matching capabilities together. This enables us to connect skills to people and to their relationships to jobs, opportunities, projects, and much more. Think of skills cloud as the flywheel that other skills-related features spring from.
With the skills cloud, we provide a maintenance free, always up-to-date universal skills language that shares insights into the capabilities of any workforce. It also verifies and measures the strength of the skills employees add to their profiles while highlighting where skills gaps are, and helps power the processes that fill those gaps. Skills cloud also fuels:
- Talent marketplace: A talent mobility platform that allows organizations to identify and access talent, while enabling workers to identify growth opportunities based on machine-learning driven skills matching for projects, gigs, and new roles.
- Skills miner: As many are painfully aware, skills are not always entered, or entered accurately, by workers. To solve this problem, we created skills miner. Skills miner can scan structured and unstructured data to automatically collect skills from things like a resume or LinkedIn profile, classifying them using a canonical list of 55,000 skills we identified from the relationship mapping described above. Using this capability, we’ve seen, on average, an increase from 25 percent of profiles listing skills to over 79 percent of profiles listing skills, making it much easier for our customers to understand the skills trends, patterns, and gaps in their workforce. In other words, they can derive the skills hidden in candidates, employees, job reqs, or learning content, and then match people to opportunities.
- Skills insights: Features a dashboard that enables customers to better understand the current strengths and gaps across their organization to determine if they have the right skills in place to execute on their business strategy. Currently in use by three global early adopter customers, it proactively collects skills from multiple sources to provide a real-time and dynamic skills footprint for a better understanding of the talent pool within any organization.
A facet of skills insights, opportunity matching, is currently in use by Patagonia to better understand, find, and match talent. As Kylie Riley, Workday manager at Patagonia, explains, “It’s really important that we have internal opportunities for our employees, because we want to make sure that they can continue to contribute in such amazing ways. We couldn’t collect skill data on every single person in the company. Our skill and data entry from our employee population was about 28 percent. With machine learning, we were able to bring that up to 73 percent. The end result of implementing machine learning is that you have real-time data and real-time information, so that you can support your people”
Shannon Ellis, director of human resources at Patagonia, added, “With Workday, our recruiters have the ability to do searching in the moment and have a greater sense of who’s out there and who’s right for the role at hand. When I first started at Patagonia, it was pretty small and you knew who everybody was. Our hope is that with Workday’s machine learning, we’re going to make that world feel small again.”
And, in the future, we have a number of skills- and career-related enhancements we’re working on. Here are a few to give a flavor of what we’re developing:
- Skills signature: Lets workers know which skills are being discovered by Workday and lets them refine the suggestions to provide better recommendations and matching.
- Skills verification: Leverages machine learning techniques to validate that the skills identified to a worker are skills they truly have.
- Career hub: A centralized space where employees can leverage the tools and resources they need to develop and grow in their careers.
- Project staffing recommendations: Recommends specific people for project assignments based on their skills, availability, and prior experience with similar projects.
It’s All About an Empowered Experience
All of these machine learning capabilities center around the user, and help to create business value for organizations. Another facet of this is personalization, and making sure each customer’s Workday experience is catered to their needs. As my colleague Stuart Bowness explains in more detail, machine learning powers our new Workday People Experience, which predicts what people want and gives them quick access to it, eliminating needless navigation with a simpler, more engaging digital experience, We believe that saving users’ time while making them more effective is one of the most valuable things we can do.
We have many more exciting machine learning enhancements to come, but I want to emphasize that we see machine learning as a tool to help everyone in an organization in ways large and small. Ultimately, empowering better business outcomes and creating undeniable value in a quickly changing world is our goal, and we look forward to continuing this journey.