The answer lies in what the interaction is designed to achieve, not just whether the question gets answered, but whether the person in front of it can complete the work they came to do.
Phase 2 of this series explored what needs to be in place before AI can be introduced successfully, across systems, use cases, data, and the organisation itself. Phase 3 shifts the focus to design: how AI experiences are constructed within Oracle HCM and what separates an experience that genuinely helps people get work done from one that simply answers questions.
This article begins with a simple distinction: helping someone find information is not the same as helping them complete a task. Understanding that difference is the foundation for designing effective AI experiences within Oracle HCM.
Information vs Action
Most AI agents can answer questions. Far fewer help people get work done.
A general-purpose knowledge assistant succeeds when it retrieves the correct information and presents it clearly. That is a meaningful capability, but it is also a narrow one. In Oracle HCM, the more valuable design question is not whether an agent can answer a question correctly. It is whether the agent can recognise where the employee, manager, or HR specialist sits within a process, and guide them into the right next step.
This distinction is easy to underestimate, because both interactions can look identical on the surface. A user types a question. An agent responds. Sometimes the interaction ends with an answer. Sometimes it helps the user take the next step. Although those experiences may appear similar, they are designed for very different outcomes.
In both cases the information is correct, but the second interaction helps connect information with the work the employee may wish to complete next. The difference is not the quality of the answer; it is how closely the experience supports the outcome the user is ultimately trying to achieve.
Phase 3 · Figure 01
Information vs Action
Information
- Answers questions
- Explains policies
- Provides guidance
- Delivers understanding
Action
- Starts processes
- Initiates transactions
- Supports decisions
- Completes work
Designing an HR agent means designing the move from one column to the other.
Oracle HCM scenario
An employee asks the Leave and Absence Analyst how much annual leave they have remaining. A knowledge-focused experience retrieves the correct balance and presents it. The question has been answered and the interaction can end there.
Other HR experiences may go further. The employee may ask whether they already have leave booked, whether they are eligible for a particular absence type, or how to request time off during a specific period. Rather than stopping at information, the experience can help the employee navigate towards the relevant leave request process by providing a deep link or guiding them to the next step.
In both cases the information is correct, but the second interaction helps connect information with the work the employee may wish to complete next. The difference is not the quality of the answer; it is how closely the experience supports the outcome the user is ultimately trying to achieve.
This is what makes HR AI design different. A knowledge assistant is judged by the accuracy of its answers. In HR, answering the question is often only part of what the user needs. The challenge is not only providing the right information. It is helping users move from understanding to action.
Four Levels of HR AI Interaction
When designing AI-enabled HR experiences, interactions often fall into four increasingly consequential categories. Whether delivered through Oracle's AI capabilities, configured experiences, or future agent designs, each level carries a different degree of responsibility and risk.
The four levels are not intended to describe a single Oracle AI agent. Rather, they describe how HR interactions can evolve from information towards workforce outcomes. Some Oracle AI capabilities focus primarily on guidance and information retrieval, while others support specific transactions and workforce processes. The closer an interaction moves towards a workforce outcome, the more significant the consequences of getting the interaction wrong become.
Phase 3 · Figure 02
The Four-Stage Shift
Increasing design responsibility
Policy Guidance
Explains policies and rules
Process Support
Explains what happens next
Transaction Enablement
Launches the appropriate process
Workforce Outcome
Supports real workforce changes
Each level carries more design responsibility than the one below it.
Policy Guidance
Explaining what a policy says and how it applies to the user, such as parental leave entitlement, relocation eligibility, working hours rules, or other organisation-specific policies and guidance.
Process Support
Helping a user understand where they are in a process, what steps remain, who may be involved, and what is likely to happen next before a decision or action can be completed.
Transaction Enablement
Helping a user move from understanding to execution by launching the appropriate Redwood experience, surfacing the correct process, or reducing the effort required to complete a task.
Workforce Outcome
Supporting activities that result in a genuine workforce outcome, such as a transfer, change to working pattern, role move, or update to employment-related information.
Oracle HCM scenario
The Personal Information Assistant (now Person Data Assistant) provides a useful example of how HR interactions can progress across the four levels.
An employee may begin by asking which personal information they are allowed to update themselves, such as contact details, emergency contacts, or demographic information. This is policy guidance, helping the user understand what is permitted and available within the system.
The conversation may then move to process support, with the employee asking how to update their marital status or where to make a change. The assistant can explain the process and direct the user to the appropriate area of Oracle HCM.
At the transaction enablement stage, the assistant helps the user move from understanding to execution by guiding them through the update process or directing them to the appropriate page or experience. Depending on the configuration and capability available, this may include deep links or conversational assistance.
The interaction ultimately results in a workforce outcome. For example, a marital status, phone number, or email address may be updated and, where approval rules are configured, submitted into the appropriate approval process.
This example illustrates how HR interactions can move from information towards a workforce outcome, even though different capabilities and processes may be involved along the way.
Providing information and supporting workforce activities are not the same thing. The further an interaction moves towards a real workforce outcome, the more important design, governance, security and business controls become. Each level carries greater responsibility than the one before it. Incorrect policy guidance may create confusion, while errors within workforce processes have a direct impact on employees, managers and organisational records.
The Platform Behind the Agent
What makes this progression particularly significant within Oracle HCM is that the agent is operating inside the same platform that already holds the workforce record. Unlike standalone AI assistants that rely primarily on documents and knowledge sources, Oracle HCM agents can draw upon organisational structures, role definitions, security controls, approval hierarchies, and business processes already present within the application.
This changes the nature of the design challenge. The goal is not simply to provide information, but to determine when and how information should be connected to an appropriate business process.
As Oracle AI capabilities continue to evolve, the distinction between answering a question and supporting a workforce action becomes increasingly important. A well-designed agent should help users move efficiently from information to action while remaining aligned with the processes already established within the organisation.
Why This Distinction Matters for Design, Not Just Capability
It is tempting to treat this as a technical question. Can the agent technically initiate a transaction, yes or no? In practice, the more important question is a design one: should this agent move from information to action in this specific context, and what does it need to know, confirm, and respect before it does?
An agent that moves too readily from answering to acting creates risk, initiating a process before the user has confirmed they actually want it, or before the right approvals and context are in place. An agent that never moves from answering to acting creates frustration, leaving the user to do manually what the agent clearly had enough information to start. The right design sits between those two failure modes, and where that line sits depends entirely on the sensitivity of the workforce action involved.
This is also where HR agent design differs most clearly from a general-purpose knowledge assistant. A knowledge assistant succeeds when it provides accurate information. HR experiences often need to go further by helping users navigate processes, complete transactions, and reach an intended outcome. As a result, the design challenge extends beyond answer quality and into how information is connected to action.
Designing an Oracle HCM AI agent is not the same exercise as designing a knowledge assistant, because the destination is different. A knowledge assistant succeeds by being accurate. An HR agent succeeds by recognising when accuracy is enough and when the person in front of it actually needs to be moved into action.
That distinction, between policy guidance, process support, transaction enablement, and workforce outcome, is the foundation the rest of Phase 3 builds on. Article 8 examines what makes that foundation effective: the workforce context Oracle HCM already holds about every user before a conversation begins, and why that context, not language generation, is where the real design value sits.
Phase 3 · Figure 03
Designing for the gap between information and action
Information
- Answers
- Knowledge
- Accuracy
AI agent design
- Context
- Governance
- Security
- Trust
- Confirmation
Action
- Processes
- Transactions
- Workforce changes
“The value of an HR AI agent is determined not by what it knows, but by how effectively it helps people move from understanding to action.”
In the Next Post
In the next article, Designing AI Agents: Applying Workforce Context, the focus shifts to what makes the same Oracle HCM agent produce dramatically different value for different users.
The answer is not language. It is context.
Oracle AI Agents Discussed in This Article
