First, a distinction worth making: AI and ML are not the same thing.
Most people use artificial intelligence (AI) and machine learning (ML) interchangeably. That distinction doesn’t matter much in everyday conversation, but inside Workday, it does, because the two things have been in the platform for different lengths of time and do different things.
Machine learning, or ML, is the older of the two. It works by identifying patterns in large amounts of data and using those patterns to make predictions or recommendations. Workday has been using ML for years. When the platform suggests a candidate for a role based on their skills profile, flags an unusual payroll entry, or predicts which employees are at risk of leaving, that’s ML at work. It’s powerful, it’s embedded in the platform, and most organisations already have access to it without necessarily knowing it.
Artificial intelligence, and specifically generative AI, is newer and more visible. Where ML predicts and recommends, generative AI creates. It can write a job description, summarise a help article, answer a question in plain language, or generate a report narrative. This is what most people mean when they talk about AI today, and it’s the area where Workday has invested most heavily in the last two years.
The reason the distinction matters is practical. Some of what Workday calls AI is really ML that has been in your subscription for years. Some of it is newer generative AI capabilities that may require additional steps to activate. Knowing which is which helps you understand what you already have and what you still need to do.
The platform: what Workday AI is
Workday AI is the umbrella term for all artificial intelligence and machine learning capabilities across the Workday platform. It covers everything from the predictive features that have been in the system for years to the newer generative AI tools being rolled out now.
Until recently, Workday used the brand name Illuminate to describe its next generation of AI. You will still see this term in Workday documentation and communications, and it remains the name for the AI platform layer that powers the features described in this article. It is worth knowing what it refers to, because it comes up regularly.
More recently, Workday acquired a company called
Sana and integrated it into the platform as the new primary way users interact with Workday AI. Sana is significant enough to deserve its own glossary entry below.
Workday AI glossary: the terms you need to know
Below is a glossary of the key terms you are likely to encounter as a Workday customer thinking about AI. We have grouped them into three areas: the underlying concepts, the platform and its components, and the commercial model.
The concepts
Generative AI
Generative AI refers to AI that can produce new content rather than simply analysing existing data. In a Workday context, this means things like generating job descriptions, drafting communications, summarising documents, and answering questions in natural language. It is the type of AI most people are thinking of when they talk about AI today.
Predictive AI
Predictive AI uses historical data to forecast what is likely to happen next. In Workday, this includes things like predicting attrition risk, forecasting cash flow, and identifying anomalies in payroll or journal entries. Much of this has been in the platform for years under the machine learning label.
Machine learning (ML)
A specific type of AI that learns from patterns in data over time. Rather than being programmed with explicit rules, an ML model improves as it processes more information. Workday has extensive ML capability embedded across HR and Finance that many organisations have not fully activated.
Agentic AI
This is where AI moves beyond answering questions or making recommendations and starts taking action. An AI agent can plan, reason, and execute multi-step tasks on its own, within defined boundaries set by the organisation. Rather than a user asking a question and getting an answer, an agent can receive a goal and work out how to achieve it, coordinating across systems and processes along the way. This is the direction Workday is moving in most actively right now.
Natural language processing (NLP)
The technology that allows AI to understand and respond to human language rather than structured commands or menus. When a user types a question into Workday Assistant and gets a relevant answer, NLP is doing the work of understanding what was asked. It is what makes conversational AI feel like a conversation rather than a search engine.
Illuminate
Illuminate is Workday’s name for its AI platform layer, announced at Workday Rising 2024. Illuminate is not a single product or feature but the underlying framework that powers Workday’s AI capabilities, including the service agents described below. You will see it referenced in Workday’s own materials. Think of it as the engine rather than the car.
Sana
Workday announced it had acquired Sana, a Swedish AI company, in early November 2025. Sana is now the primary interface through which users interact with Workday AI. Rather than navigating through Workday’s traditional menus and workflows, users can interact with Workday conversationally through the Sana workspace, asking questions, running reports, and triggering actions in plain language.
There are two versions relevant to Workday customers. Sana Core is included for organisations that have signed the uMSA (see below) and gives users conversational access to Workday HR and Finance data, read and write actions through the Workday Self-Service Agent, and automated workflows for HCM and Finance processes. Sana Enterprise is an upgraded tier that extends this capability beyond Workday into other enterprise systems, including Gmail, Salesforce, SharePoint, Slack, and ServiceNow.
Workday Assistant
The conversational AI tool built into the Workday platform. Workday Assistant allows users to ask questions in plain language and get relevant, personalised answers based on their role and location. Workday Assistant is being retired in September 2027 and replaced by the Self-Service Agent. If you enable the Self-Service Agent now, Workday Assistant is automatically disabled. The two cannot run simultaneously.
Self-Service Agent
The Self-Service Agent is a significant upgrade to Workday Assistant, and the primary conversational AI agent for most Workday customers. It handles the full range of employee and manager self-service interactions, answering questions about pay, benefits, time off, and HR policies, and enabling employees to act on their own data directly through conversation.
It covers core HCM, Workforce Management, Talent Management, Payroll, and Help. Sana is the experience layer. The Self-Service Agent is the Workday intelligence behind it. The two work together but are distinct things.
Service agents
Service agents are AI agents built by Workday to handle specific business processes. Each agent focuses on a defined area of work. The Recruiting Agent automates sourcing, outreach, and interview scheduling. The Expenses Agent creates and submits expense reports. The Succession Agent prompts and automates succession planning. The Self-Service Agent handles employee queries and HR tasks directly. The Financial Audit Agent scans financials continuously for errors and compliance issues.
Unlike features that surface information or make recommendations, agents take action within boundaries defined by the organisation. Some agents are currently in early adopter phases, with broader availability being rolled out over time.
Agent System of Record (ASOR)
A central hub where administrators can register, configure, activate, and deactivate agents, whether built by Workday, a partner, or the organisation itself. It provides visibility into how agents are performing and what actions they are taking.
Agent Gateway
The technical layer that controls how AI agents interact with Workday data and with each other. It is what makes it possible for agents to work across systems securely and consistently. Most HR and Finance leaders will not configure this directly but will encounter the term in technical conversations about activation.
Workday Extend
The platform that allows organisations and partners to build custom applications and integrations on top of Workday. Relevant in the AI context because it is the route through which some custom agent-building capability is accessed.
Skills Cloud
Workday’s underlying skills ontology, which maps relationships between skills, roles, and capabilities across the workforce. It powers AI features in talent management, including skills suggestions, learning recommendations, and internal mobility tools. For organisations using Workday for workforce planning, Skills Cloud is the data foundation that makes AI recommendations relevant rather than generic.
Intelligent Answers
An AI feature within Workday Help that surfaces relevant answers from an organisation’s help article library when employees search for information. Rather than raising a support ticket, an employee searching for a policy answer is shown the most relevant content automatically. It reduces administrative overhead and improves the employee experience. This is an example of ML-powered AI that is already active for many organisations without them necessarily knowing it.
HiredScore
Workday acquired HiredScore, an AI-powered talent intelligence company, in 2024. HiredScore brings dedicated AI capability for recruiting and internal mobility, including candidate scoring and opportunity matching. Where Workday’s existing recruiting features manage the process, HiredScore adds a layer of intelligence that helps organisations prioritise and act on the right talent at the right time. It is integrated into the Workday platform rather than sitting alongside it as a separate product.
Workday Build
Workday Build is Workday’s unified platform for creating, extending, and integrating business solutions. It brings together low‑code app development, integrations, analytics, and AI tools, all running on Workday’s secure, unified data model. It also includes developer programs, learning paths, and a marketplace ecosystem that helps customers and partners build, share, and deploy Workday‑native solutions.
Workday Flowise Agent Builder
Workday’s platform for building custom AI agents and applications. Flowise Agent Builder is a low-code tool within Workday Build that allows organisations to create their own agents tailored to their specific processes. It is available to customers on Workday Extend Professional. For most organisations, this is a future consideration rather than an immediate priority, but it signals Workday’s intention to make agent-building accessible beyond technical teams. Flowise is aimed at helping partners and developers familiarise themselves with the technology by allowing them to build and test on the platform directly.
Workday Orchestrate
Workday Orchestrate is a platform designed for creating, managing, and sharing integrations between Workday and external systems. It supports high volumes of data and complex integration patterns while ensuring security and scalability. The integrations are built using the Orchestration Build, which provides a drag-and-drop development environment that allows users to create workflows (orchestrations) without coding. Integrations created using Workday Orchestrate are referred to as integration apps, which can automate business processes and facilitate real-time API interactions.
The commercial model
uMSA (Universal Main Subscription Agreement)
Workday’s updated contractual framework, introduced in 2024 to replace the legacy Main Service Agreement and the previous Innovation Services structure. It is the gateway to Workday’s AI and agent capability. Organisations on the legacy MSA need to migrate to the uMSA to access service agents in production.
Migration happens at renewal, or organisations can sign an Early Migration Amendment outside of a renewal cycle. If you are not sure which agreement your organisation is on, you can check through your Workday Community profile. It is usually the first conversation worth having.
Flex Credits
Flex Credits are Workday’s consumption-based model for AI and platform features. Each agent and platform capability is assigned a credit rate based on its operating cost and value. Organisations receive a base allocation of credits as part of their subscription, renewed annually, and can purchase additional credits as usage grows.
The model is designed to give organisations flexibility to adopt AI at their own pace rather than committing to large upfront purchases. Credits are fungible, meaning they can be applied across different agents and platform features rather than being locked to a specific use case.
The practical thing to understand is that Flex Credits are not free. The base allocation is included, but meaningful use of agents will likely require additional credits. The rate card is not publicly available, which makes budgeting harder than it should be. Understanding your likely consumption before you start activating is worth doing, and it is something a Workday partner like Preos can help with.
When a service agent completes an action in Workday, such as retrieving pay information, updating a record, or processing a request, it does so by calling on a specific tool within the platform. Each of these interactions is a tool call. Flex Credits are consumed per successful tool call, not per conversation. A conversation that does not result in any action uses no credits at all.
SKU
A SKU is Workday’s term for a purchasable product or feature set. When someone refers to whether a capability is included in your SKU, they are asking whether it is part of what you have already paid for. Some AI features are included in existing SKUs. Others, like Sana Enterprise, require an additional SKU purchase.
