The annual Workday Rising conference returned to its home ground of San Francisco this week, after an absence of several years. Founder and co-CEO Aneel Bhusri kicked off yesterday’s product innovation keynote in an assured double act with co-President and tech wizard Sayan Chakraborty, where AI, inevitably, was the main focus of attention. But this was not your everyday generative AI that just scoops everything off the Internet, mixes it up to a secret recipe and then serves up a potentially toxic result. No, this was Workday AI. As Chakraborty puts it:
Honestly, you can’t be semi-accurate when you’re running payroll. You need payroll to be correct — there’s not a lot of room for hallucinations in payroll. You can’t be semi-accurate when you’re adding to your general ledger or trying to do a financial close.
We treat our AI models the same way we treat everything else at Workday — with precision, with accuracy. And we require high-quality data, so we can generate high-quality content and high-quality recommendations.
In yesterday’s announcements, Workday plans to harness generative AI over the next six to 12 months to be able to automatically draft a whole range of enterprise content, from job descriptions and knowledge articles to employee growth plans, statements of work and collections letters. Another use case is to analyze contracts to discover discrepancies between the signed contract and what’s recorded in the CRM and financials system, propose corrections and then confirm when the issue has been fixed.
New AI capabilities are also coming to the Workday Extend developer platform, with an AI gateway that will give developers access to Workday AI and ML services, including skills analysis based on Workday Skills Cloud, sentiment analysis, intelligent document analysis to extract business-relevant data, and a time series forecaster. Workday Extend will also natively support several AWS AI services, including image processing and language translation.
The low-code Workday App Builder is gaining a developer co-pilot feature, which creates application code from natural language prompts, along with other low-code and no-code capabilities. Workday is also looking into the use of generative AI to extend the natural language capabilities of a conversational user interface.
A final announcement, which Bhusri and Chakraborty saved for the end of the keynote as ‘one more thing’, is the launch of an AI marketplace of certified third-party apps and services that can be accessed in Extend via the AI Gateway. Set for availability at latest in Q2 next year, the first wave of 15 early adopter partners was unveiled yesterday, including consulting partners such as Accenture and Kainos, ISVs such as Sana and Vertex, and AI startups such as Hiredscore and Paradox.ai.
Emphasizing Workday’s AI track record
In the keynote, Bhusri emphasized Workday’s long track record of building AI into its platform, walking the audience through a timeline from the company’s foundation in 2005 to its first experiments with AI in 2013, and the introduction of its first ML-enabled product features in 2015. He noted that Workday started working on Large Language Models (LLMs) in 2020, long before the current wave of excitement around generative AI. Out of that work comes the domain-specific LLMs that power the latest announcements. Chakraborty takes up the story:
What we use at Workday is what we refer to as enterprise large language models, or enterprise LLMs. These rely on high-quality data. This is data that already has strong regulatory privacy, security, and IP ownership built in. So when we’re training Workday models, we know where the data comes from. We know who owns it. We know where there may be concerns about specific use cases. We can use methods like grounding, to ensure our models provide truth, and allow you to tap into these benefits of LLMs without really exposing you and your employees to the dangers of internet-trained LLMs.
Safety and trust form the basis for realizing the full benefits of the technology. He continues:
If we can enable generative AI safely, it does open up fantastic new opportunities. It can allow you and your employees to be your best selves at work. You can spend less time working through mundane tasks and getting to the work you actually want and need to do. Like with any new technology, like any new technology transition, this is a little over-hyped. But underneath the hype, there’s real value here, and we are committed to delivering that value to you.
Other announcements yesterday brought new capabilities to Workday HCM, many of them AI-assisted:
- A new Manager Insights Hub uses AI and ML to surface personalized recommendations for employee career development, such as suggesting connections, mentors, and gigs based on their skills interests.
- A new Flex Teams capability helps to rapidly assemble teams for specific tasks and projects. This uses AI within Workday Skills Cloud to quickly identify suitable talent from across the organization, assemble a team, and define roles.
- A new Home and Insights feature delivers curated insights from Workday to provide managers with a holistic view of relevant information related to their teams, including important dates such as birthdays, anniversaries and time off. A My Tasks section on the home page enables quick actions to complete tasks, and these capabilities are also integrated into Microsoft Teams and Slack.
- A new UI integration to Adaptive Planning brings workforce planning directly into the Workday HCM user interface. Other new capabilities here include automated headcount reconciliation and a new planning configuration manager that simplifies setup and maintenance.
Adaptive Planning also gets an upgrade to its AI-assisted intelligent modeling engine, bringing faster dashboard performance and OfficeConnect reports, a predictive forecaster, and the ability to create personal what-if scenarios and automated report scheduling and distribution.
After all the hype around generative AI earlier this year, enterprise reality hits home as we enter the fall conference season. Despite the undoubted productivity potential of harnessing large language models, organizations aren’t going to entrust their core operations and customer outcomes to the technology unless they’re absolutely confident it won’t go awry. Workday hit all the right notes in its product keynote here, emphasizing its track record and all the ways in which it is able to corral this wayward technology to make sure it can become a reliable servant. The proof of the pudding of course comes when these capabilities go into production with customers. But Workday has started out on the right foot.