Introduction

Workday held its Rising conference this week in San Francisco, CA. As expected, Artificial Intelligence (AI), and Generative AI (Gen AI) in particular, was prominent in keynotes and the announcements. During Fall conference season, all vendors look to stand out. Workday is no different. With most (probably all) vendors touting new Gen AI capabilities, that can be hard. Workday’s focus on building its own what-they-call “Enterprise Large Language Models” as well as an AI partner ecosystem (see announcement of its new AI Marketplace) were a different take from what we have seen thus far from other Work Tech vendors. Though Gen AI was a major focus for Workday, it also announced many other improvements to their business applications and development / extension capabilities (Workday Extend).

Building AI into the Platform

Like many vendors, Workday is not new to AI, even Gen AI, having started to work with Gen AI in 2020. Workday is a proponent of what they call Ethical AI and ML (Machine Learning) Trust. The blog post lays out three goals for responsible and trustworthy AI:

Amplify Human Potential: Put people at the center and have a “human in the loop”
Positively Impact Society: Develop solutions that align with Workday’s values
Champion Transparency and Fairness: Clear documentation of how models are built and work. Risk-based review for unintended consequences. Innovation in algorithmic fairness and explainability.

This led Workday to build its own Large Language Models (LLMs) which are trained on its own curated datasets. This is what Workday calls Enterprise LLMs. We think this is a good approach that balances the benefits and potential of Gen AI while addressing Enterprise governance and risk concerns. It also sets the stage for improvements across Workday.

Application Improvements

Workday is taking a pragmatic approach to leveraging Gen AI capabilities in its applications. Gen AI is good at generating text (new text, creating summaries, etc.). Workday identified several use cases that take advantage of text generation including job descriptions, knowledge base articles, employee growth plans, and statements of work (for professional services). These are beneficial use cases but also require appropriate guardrails for bias (job descriptions and career growth plans) and accuracy (knowledge base articles). Workday starting with these use cases shows that it is confident that has built in the necessary safeguards to get value from Gen AI, but they are not groundbreaking.

Workday also announced Gen AI improvements to Adaptive Planning to streamline headcount planning as well as a new solution, Manager Insights Hub, that makes it easier for managers to find and act on insights discovered via AI. Workday customers struggling to get managers to use (and act on) reports and dashboards should investigate Manager Insights Hub to drive stronger adoption.

Workday is updating its Digital Assistant to use Gen AI capabilities. The most interesting use case is natural language analytics. It can respond to a question with charts and visualizations to answer the questions (and allow follow up questions). This shows some of the true power of Gen AI, enabled by previous investments in Workday Query Language (WQL). Though not the only vendor to show this type of capability, it is a more advanced use case and a better glimpse into the art of the possible.

Manager Insights Hub represents “pushing” Workday insights to managers. The Digital Assistant update shows how managers can “pull” insights more easily from Workday. Employee enablement and experience was also highlighted with the introduction of the User Success Platform, which provides contextual guidance on how to complete complex tasks in Workday. For customers wanting to increase employee and manager adoption in Workday, these are important new capabilities to consider.

Developer (Workday Extend) Improvements

Workday has been building out its Extend solution for almost 7 years. Today, there are more than 5,000 developers who have built more than 1,300 applications. Workday, recognizing the importance of apps and extensions built on Extend having access to AI and ML services, announced AI Gateway. Workday plans to offer the following services through AI Gateway:

Skills Analysis (using Workday Skills Cloud)
Sentiment Analysis (analyze text for sentiment)
Document Intelligence (extract data from business documents)
ML Forecaster (time series forecasting for seasonal and non-seasonal predictions)

These enhancements, included with Extend Professional, expand the range of applications that can be built on the Workday platform. Workday also announced an expanded partnership with Amazon Web Services (AWS) that will allow Extend developer to tap into AWS services via connectors. This will also increase the ability of Extend developers to expand their use cases for Workday apps and extensions.

Workday is planning to add Gen AI to App Builder to improve the developer experience via Developer Copilot (planned in the next 6-12 months). Developer Copilot will turn natural language into application code. Developer Copilot along with other low-code / no-code enhancements announced for App Builder should increase developer productivity and make Workday Extend easier to adopt.

AI Marketplace (in beta now, expected to be live in the first half of 2024) is the first step toward developing a partner ecosystem that enables Workday customers to leverage AI innovations happening beyond Workday with data from their Workday platform. We think this is a major step because there is so much innovation and investment happening with AI and ML that one vendor cannot realistically do it all themselves. The 15 early adopters represent Services Partners (Accenture and Kainos), Hyperscalers (AWS, Google Vertex AI), Workday Ventures Partner Companies (Auditoria, BetterUp, Paradox.ai, Sanalabs, and WorkBoard), and Software Partners (Avalara, HiredScore, TechWolf). Workday customers who are leveraging or thinking of leveraging these partners should keep a close eye on how they can leverage the AI Marketplace going forward.

Conclusion