Back to Oracle HCMA woman stands in a dark gallery beside a glowing red network of connected points.

ORACLE HCM · AGENTIC APPLICATIONS

Agentic applications for HR

Oracle's AI that coordinates whole pieces of HR work, from surfacing information to carrying out approved actions — not just single tasks.What you'd be asked to review or approve, how much each application does without you, and what an illustrative run looks like from start to finish.

SHARE THIS PAGE

HOW TO USE THIS PAGE

See what Oracle's agentic applications mean for your role, and how much each one does without a person.
  1. 1

    Pick your role to see which applications would prepare work for you.

  2. 2

    Use the ladder to see how much each one does on its own. Select an application for Oracle's own words.

  3. 3

    Step through one run to see where a person stays in charge, then work through the checklist.

What's an agentic application?

AI agent

Does a defined task: answers a policy question, summarises a check-in or updates a record. Most of Oracle's AI in HCM since 25C is in this group.

See every HCM AI agent →

Agentic application

Works towards an outcome, such as a workforce plan or a payroll run that's ready to go. It can bring together several agents, data sources and steps, deciding which to use and in what order — though how much it does on its own varies, from surfacing information to carrying out approved actions. Oracle gives many of them names like workspaces and command centres, though not all.

Fusion Claw

Oracle's "governed agentic execution runtime" for its newer agentic applications, announced in September 2026. It's built for larger, longer-running work, and it records what each run did in an "Outcome Receipt".

ANNOUNCED — NO DATE OR PRICING YET

Fusion Companion and Fusion Agentic Home

Oracle announced both at Oracle AI World in October 2026. Fusion Agentic Home is a personal start page in Fusion where people set objectives, hand work to agents and track it to completion. Fusion Companion lets people do the same through supported AI tools such as ChatGPT, using their own permissions. Oracle hasn't given dates or pricing yet.

Read Oracle's announcement ↗

1 · WHAT YOU'D REVIEW

Pick your role

An AI agent does one task; an agentic application coordinates several agents to prepare a whole piece of work for you to check. See which of Oracle's agentic applications are built for your role, and whether each one gives you work to review, suggestions to act on, or help you use yourself.

Every agentic application Oracle has announced for HR, with what each one would hand you. Some applications are counted under more than one role, so the role totals add up to more than the full list.

READY FOR YOU TO REVIEW AS

Workforce plans weighing development, redeployment, hiring and external talent, ready for your organisation's review.

Does this much on its own:

Rung 2: Recommends

READY FOR YOU TO REVIEW AS

Staffing plans that balance skills, schedules, labour rules and costs.

Does this much on its own:

Rung 3: Prepares for approval

READY FOR YOU TO REVIEW AS

Position conflicts and downstream risks, with suggested corrections to approve.

Does this much on its own:

Rung 3: Prepares for approval

READY FOR YOU TO REVIEW AS

Scheduling, time and absence risks, with guidance on coverage decisions.

Does this much on its own:

Rung 2: Recommends

READY FOR YOU TO REVIEW AS

Grouped candidates and outreach campaigns to approve within your recruiting process.

Does this much on its own:

Rung 4: Carries out approved actions

READY FOR YOU TO REVIEW AS

Candidates ready to advance, and next steps to approve.

Does this much on its own:

Rung 3: Prepares for approval

READY FOR YOU TO REVIEW AS

Ranked candidates, scheduled interviews and stalled requisitions.

Does this much on its own:

Rung 2: Recommends

READY FOR YOU TO REVIEW AS

Pay, performance, talent and absence signals, with one-click actions to approve.

Does this much on its own:

Rung 3: Prepares for approval

READY FOR YOU TO REVIEW AS

Suggested rating changes for your team before a talent review, with explanations.

Does this much on its own:

Rung 2: Recommends

READY FOR YOU TO REVIEW AS

Coaching insights and recommendations for your team.

Does this much on its own:

Rung 2: Recommends

READY FOR YOU TO REVIEW AS

Your team's learning, with compliance risks and skill gaps flagged early.

Does this much on its own:

Rung 1: Shows and flags

READY FOR YOU TO REVIEW AS

Compensation plans within budget and pay guidelines, for review and approval.

Does this much on its own:

Rung 3: Prepares for approval

READY FOR YOU TO REVIEW AS

Likely payroll problems and their causes, with follow-up actions to review.

Does this much on its own:

Rung 3: Prepares for approval

READY FOR YOU TO REVIEW AS

Significant differences from earlier periods, with follow-up actions to approve.

Does this much on its own:

Rung 3: Prepares for approval

READY FOR YOU TO REVIEW AS

Enrollment and plan-use trends, life-event and court-order issues, and pending actions or documents, with follow-up actions for administrators.

Does this much on its own:

Rung 1: Shows and flags

READY FOR YOU TO REVIEW AS

Open roles, the skills they need, and suggested training and gigs.

Does this much on its own:

Rung 2: Recommends

READY FOR YOU TO REVIEW AS

Your requests, follow-ups and answers in one place.

Does this much on its own:

Rung 1: Shows and flags

READY FOR YOU TO REVIEW AS

Learning guidance for your next steps.

Does this much on its own:

Rung 2: Recommends

READY FOR YOU TO REVIEW AS

A guided course path, practice and assessment feedback, with progress and learning outcomes recorded in Oracle Learning.

Does this much on its own:

Rung 1: Shows and flags

READY FOR YOU TO REVIEW AS

Priority workforce activity, staffing and span-of-control risks, compliance records needing follow-up, and contextual summaries or draft review notes.

Does this much on its own:

Rung 1: Shows and flags

READY FOR YOU TO REVIEW AS

A draft self-paced course with an editable structure and learner-facing material, ready for a specialist to refine, preview and activate.

Does this much on its own:

Rung 3: Prepares for approval

READY FOR YOU TO REVIEW AS

Contracts analysed, with suggested next steps.

Does this much on its own:

Rung 2: Recommends

2 · HOW MUCH IT DOES ON ITS OWN

How much each one does on its own

Each application is placed on a five-rung ladder of my own, based on Oracle's own description of what it does — this isn't an Oracle classification or a statement of the product's default settings. Select one to see Oracle's words. A red ring shows a Fusion Claw application: Oracle says customers choose the level of automation, up to full auto within delegated authority.

RUNG 1Shows and flags

Brings information together and flags what needs attention.

RUNG 2Recommends

Suggests what to do. A person acts, often somewhere else.

RUNG 3Prepares for approval

Prepares the action. A person approves it before anything happens.

IN ORACLE'S WORDS · PAYROLL COMMAND CENTER

“... highlighting issues and potential causes; and preparing recommended follow-up actions for review.”

Fusion Claw application: Oracle says you can set it up to full auto.

Placed on rung 3 from Oracle's description.

Read Oracle's announcement ↗

RUNG 4Carries out approved actions

Once a person approves, it does the work itself.

RUNG 5Runs within delegated authority

Acts on its own inside limits you set. Oracle calls this full auto.

No applications are placed at rung 5 in this guide. Oracle says the Claw-powered ones can be set this far.

IN ORACLE'S WORDS · PAYROLL COMMAND CENTER

“... highlighting issues and potential causes; and preparing recommended follow-up actions for review.”

Fusion Claw application: Oracle says you can set it up to full auto.

Placed on rung 3 from Oracle's description.

Read Oracle's announcement ↗
Where Oracle describes itFusion Claw: can be set up to rung 5, your choice

3 · INSIDE AN ILLUSTRATIVE RUN

Payroll Command Center, step by step

An illustration built from Oracle's description, to show where a person stays in charge. The example issue is made up. Oracle's screens will look different.

STEP 1 OF 6 · Payroll manager

STEP 1 OF 6 · Payroll manager

Someone sets the goal

A payroll manager sets the outcome: this month's run ready on time, with no material errors — meaning errors large enough to affect someone's pay or compliance. With Fusion Claw, the objective sits inside limits you define: policies, permissions, risk thresholds, approval requirements and escalation boundaries.

YOUR CONTROL AT THIS STEP

Set the level of automation for this application, in writing, before the first run.

Each application also shows my risk rating, using the method on the AI HR Lifecycle Map: 7 higher risk, 14 take care, 1 lower risk. How I rated the risk →

BEFORE YOU SWITCH ONE ON

What to agree first

  • Confirm with Oracle what it costs and what your licences include.
  • Check it's available in your environment and region, and read Oracle's note for its setup.
  • Agree its level of automation, in writing.
  • Name its owner and its approver, as roles, not people.
  • Check the approver can see the recommendation, the reasons and the data behind it, and has the time to review it properly.
  • Make sure rejecting or changing a recommendation is easy and recorded.
  • Write test questions with expected results, and run them after every quarterly update.
  • Review every quarter whether approvers are actually changing anything.
  • Add it to the AI Agent Tracker.
Open the AI Agent Tracker →
BUILD, BUY OR WAIT?Read the article →When to use Oracle's application, when to build in AI Agent Studio, and when to wait.

Oracle hasn't announced agentic applications for onboarding, engagement and wellbeing, or leaving and alumni yet.

Last reviewed October 2026

Risk ratings are my own practitioner view, using the method on the AI HR Lifecycle Map. They're not legal advice or an assessment of any organisation or product. Oracle's applications, names and availability change each quarter, and pricing and licensing vary by contract. Last reviewed: October 2026.