Inventory Task Allocation Assistant
An Oracle AI Agent Studio workflow agent, delivered as a preconfigured agent template in Oracle Fusion Cloud Inventory Management (Advanced Inventory Management), that runs on a schedule and automatically assigns open, unassigned inventory tasks to qualified warehouse workers who are on shift, using Workforce Scheduling shift data, zone and activity qualifications, worker task load and worker performance metrics, subject to organization-specific rules defined in the agent prompt. Oracle documents no human approval step between the agent's evaluation and the task assignment it writes, and documents no agent-issued command to physical equipment.
Recorded characteristics
- Function
- Oracle describes the Inventory Task Allocation Assistant as a "workflow-based AI agent" that "automatically assigns open inventory tasks to qualified workers", "designed to run in the background at scheduled intervals" and helping "balance workload across the warehouse by considering shift availability, worker capacity, task priorities, and worker performance metrics". Oracle enumerates what it does at each scheduled run: it "retrieves open, unassigned inventory tasks from Workforce Scheduling-enabled zones"; "identifies organizations that are eligible for automatic task allocation based on configured inclusion and exclusion rules"; "determines active shifts that overlap with the evaluation window by using the department time zone defined in the schedule generation profile"; "retrieves workers assigned to the active shifts who are qualified for the task's zone and activity"; "evaluates each worker's current task load and performance metrics, including average task completion time and completed task volume"; "applies organization-specific rules to control task allocation behavior, including worker capacity limits, task assignment windows relative to shift start and end times, and task type prioritization"; and "assigns tasks using a balanced workload strategy, where tasks are first assigned to the worker with the fewest active tasks, and ties are resolved by selecting the worker with better performance metrics". Oracle states that when all workers reach the configured maximum, "any remaining tasks remain unassigned. During the next scheduled run, remaining tasks are assigned as workers complete existing tasks and become available", and that tasks are "dispatched automatically as workers complete existing tasks". The single documented output of the agent is therefore an assignment of an existing inventory task to a worker. Oracle does not document this agent creating, cancelling, re-prioritising, releasing or closing inventory tasks, nor monitoring task execution.
- Data access
- Documented inputs only: open, unassigned inventory tasks in Workforce Scheduling-enabled zones; task zone and activity type; task type (for task type prioritisation); inventory organizations and the configured inclusion or exclusion lists; active shifts and the department time zone from the schedule generation profile; workers assigned to those shifts and their qualification for the task's zone and activity, including competencies that can be defined for an activity type in a zone as part of the Inventory Zone definition; each worker's current task load; and worker performance metrics, specifically "average task completion time and completed task volume". Oracle's 26D entry adds, for a forthcoming release, a configured list of delivery carts used by robotic material handling equipment. Oracle does not document this agent reading item, quantity, source or destination location, inventory balances or any wider supply chain data set; those categories are recorded as not established rather than as absent.
- Actions
- Can take actions
- External actions
- Unknown
- Human confirmation
- Not required
- Permission basis
- Not established
- Administrative control
- Documented controls, all of them configuration rather than per-assignment review: the feature is delivered as a preconfigured AI Agent Studio agent template named Inventory Task Allocation Assistant, which an administrator must use or copy to create an agent team, then publish. The Enable Security Console External Application Integration (ORA_ASE_SAS_INTEGRATION_ENABLED) site-level profile option must be set to Yes to enable permission groups for roles. Allocation behaviour is governed by the Task Assignment Rules node, the first node in the agent workflow, whose Prompt field holds a predefined rule template in which lines prefixed with ## are comments. Within it administrators set Include All Organizations (Yes by default), an Organizations Inclusion list or an Organizations Exclusion list, and per-organization rules: Maximum Number of Tasks Allowed per Worker, Assignment Window Before Shift Start, Assignment Window After Shift Start and Task Type Assignment Order. Execution timing is controlled through Agent Team Settings, Triggers tab, by adding a Schedule trigger defined as "either a fixed interval or a recurring execution pattern"; Oracle advises scheduling "as frequently as needed". Worker eligibility is controlled outside the agent through Workforce Scheduling shift generation and publication and through competencies defined for an activity type in a zone in the Inventory Zone definition, where "only workers who are scheduled with the required competencies are considered for task assignment within the zone". Access is controlled through configured job roles: administration requires SCM Intelligent Agent Management Duty (ORA_RCS_SCM_AI_AGENT_MANAGEMENT_DUTY and ORA_RCS_SCM_AI_AGENT_MANAGEMENT_DUTY_HCM) and Fai Genai Agent SCM Administrator Duty (ORA_DR_FAI_GENERATIVE_AI_AGENT_SCM_ADMINISTRATOR_DUTY); interacting with agents in product pages requires Fai Genai Agent Runtime Duty (ORA_DR_FAI_GENERATIVE_AI_AGENT_RUNTIME_DUTY) plus permission groups enabled in the Security Console; and interacting with this agent specifically requires Review Inventory Task Assignments Using Responsive Inventory (INV_REVIEW_INVENTORY_TASK_ASSIGNMENTS_USING_RESPONSIVE_INVENTORY_PWA_PRIV), View Schedule Generation Profiles (HTS_VIEW_SCHEDULE_GENERATION_PROFILE_PRIV) and View Shifts (HTS_VIEW_SHIFTS_PRIV). New instance sets in data security policies control access to Workforce Schedule and Scheduling Workload Plan data for the inventory organizations a user is authorised to access. Oracle also documents that managers can "override assignments for schedules that are in progress" from the Inventory Task Assignments page, which is a human override of assignments rather than an approval gate. Oracle does not document a capability-specific pause or stop control, an allocation audit trail attributable to the agent, or a review queue for agent-made assignments.
- Default state
- Disabled
- Availability
- Oracle lists "AI Agent: Inventory Task Allocation Assistant" under Advanced Inventory Management with release "Update 26B" and type "AI agent" in its Supply Chain & Manufacturing features with AI list. In the 26B Inventory Management feature summary the feature is shown with Impact to Existing Processes "None" and Action to Enable "Setup Required"; Oracle defines the "None" impact category as "These features are not delivered enabled. These features will need steps to use. You can choose if and when to enable them." The Redwood adoption list shows the same feature as "Setup Required" with auto-enablement "Does not expire", meaning Oracle has not scheduled forced enablement. The feature requires Oracle HCM Workforce Scheduling "to be set up, with worker schedules generated, published, and integrated with Inventory Management", and requires Oracle AI Agent Studio for Fusion Applications to deploy the agent template. The 26B robotic cart feature entry directs implementers to the licensing and enablement information in the "How do I set up Advanced Inventory Management" guide and to enable Advanced Inventory Management through the Manufacturing and Supply Chain Materials Management offering, indicating that Advanced Inventory Management enablement is a prerequisite surface for this module. Oracle does not state on these pages whether the agent is generally available, preview, or restricted by region, language or environment type; those are recorded as not established. The robotic material handling extension is documented in the 26D What's New for Inventory Management, a readiness document whose revision history records it was created on 04 SEP 2026, ahead of the 26D update; it is therefore treated here as forthcoming-release documentation and not as current authority.
- Licensing
- Oracle does not state a separate or additional licence for this agent on the pages reviewed. The documented prerequisites are Oracle Fusion Cloud Inventory Management with Advanced Inventory Management enabled through the Manufacturing and Supply Chain Materials Management offering, Oracle AI Agent Studio for Fusion Applications, and Oracle HCM Workforce Scheduling. Oracle points to the licensing and enablement information at the beginning of the "How do I set up Advanced Inventory Management" guide rather than stating licence terms in the feature entry, so the precise licensing position is recorded as not established.
- External model or provider
- Not established
- Limitations and uncertainty
- Recorded boundaries and open questions: (1) Recommendation versus assignment. Oracle states this agent "automatically assigns open inventory tasks to qualified workers" and "assigns tasks using a balanced workload strategy". That is an assignment written to Oracle inventory task records, not a recommendation presented for acceptance, which is why the action capability is can_take_actions. (2) Relationship to the Inventory Tasking Assistant. These are recorded as two separate Oracle features, not one renamed capability. Oracle lists them separately in its AI feature list as "AI Agent: Inventory Tasking Assistant" (Update 26A) and "AI Agent: Inventory Task Allocation Assistant" (Update 26B), and neither entry describes the other as replacing it. Their documented behaviour also differs: the 26A Tasking Assistant "analyzes live warehouse data, operator performance, and throughput to recommend the best task-to-operator or operator-to-zone assignments", helps managers assign, unassign and reassign tasks, suggests zone deployment and lets operators self-assign, and is added to a Redwood page through a guided journey - a human-invoked, recommendation-centred agent. The 26B Task Allocation Assistant is a scheduled background workflow agent that assigns without a documented human step. No Inventory Tasking Assistant record is created here. (3) Human worker allocation is the current documented scope: assignment targets are workers on active shifts qualified for the task's zone and activity. Oracle documents the allocation factors as shift availability and active shift overlap, zone and activity qualification including zone competencies, current task load, average task completion time, completed task volume, worker capacity limits, assignment windows relative to shift start and end, and task type order. Proximity, location, certification and equipment eligibility are not documented factors for this agent and are not recorded. (4) Robotic material handling equipment. Oracle's 26D entry states that the agent "can automatically assign receipt delivery tasks to robotic material handling equipment when receipt deliveries are configured to be performed by robots", with the list of delivery carts configured in the Task Assignment Rules prompt. This is a change to the assignee of an Oracle task record, not a command to a robot, and it is documented in a 26D readiness document created on 04 SEP 2026; it is therefore recorded as forthcoming-release evidence and is deliberately excluded from the current capability state. (5) Assignment versus dispatch versus physical execution. Oracle uses the word "dispatched" only for tasks flowing to workers on subsequent scheduled runs. Physical robotic delivery is documented separately as an integration feature, "Redwood: Deliver Goods with Robotic Material Handling Equipment" (25D), in which material is picked onto a named delivery cart, "the delivery cart name defined in Oracle should match the delivery cart name in the partner system", and "when the autonomous delivery robot drops off the cart at the location, the delivery is updated to delivered" with a note reading "Delivered by Autonomous Guided Robot". The robot is operated by a partner system; Oracle records the outcome. Nothing in Oracle's documentation establishes that this AI agent sends commands to robots, to a warehouse control system, or to any other external system. (6) Physical-world boundary. The documented chain is: AI reasoning over tasks, shifts, workers and rules, then an Oracle inventory task assignment record, then execution by a warehouse worker or, in the forthcoming robotic case, by an external robotic system through a separate deterministic integration, then physical movement of goods. The agent acts only at the second stage. It does not drive robots, move inventory, operate equipment or pick or transport goods. (7) External action capability is recorded as unknown, not no: Oracle does not affirmatively state that the agent takes no action outside Fusion, and the agent reads Workforce Scheduling data that originates in Oracle HCM, a separate Fusion pillar. (8) Human confirmation is recorded as not_required for the agent's own assignment action, on the strength of Oracle's repeated "automatically assigns" language and the absence of any documented approval, review or acceptance step in the workflow; managers can override in-progress schedule assignments afterwards. This is a conclusion about the documented flow, and a future Oracle page describing a review step would change it. (9) Permission basis is recorded as not_established. Oracle documents duty roles and privileges that human users need to administer or interact with the agent, and data security instance sets governing user access to Workforce Schedule data, but does not state under whose identity or privileges the scheduled agent run itself writes task assignments. No dedicated agent identity and no agent-specific runtime role are documented. (10) Model and provider are not established: Oracle names no model or provider for this agent, and generic AI Agent Studio model availability is not attributed to it. (11) Availability detail is thin: release status beyond "Update 26B", region, language and environment constraints are not stated on the pages reviewed. (12) Not documented for this agent and therefore not recorded: creating tasks, cancelling tasks, changing task priority or status, releasing tasks, monitoring execution, reacting continuously to changing warehouse conditions between scheduled runs, and any interaction with item, quantity or location data. (13) Monitoring limitation: the primary monitored page is a release-pinned 26B What's New entry, so it captures amendments to the 26B description but will not by itself surface a later release's changes to the agent.
Evidence
- AI Agent: Inventory Task Allocation Assistant (Inventory Management 26B What's New)
Supports: Function · Actions · Data access · Human confirmation · Admin controls · Permission basis · Default state · Availability · Limitations · External actions · Primary source
States that the Inventory Task Allocation Assistant is a workflow-based AI agent that automatically assigns open inventory tasks to qualified workers, is designed to run in the background at scheduled intervals, and balances workload using shift availability, worker capacity, task priorities and worker performance metrics.
Enumerates the per-run behaviour: retrieving open unassigned inventory tasks from Workforce Scheduling-enabled zones, identifying eligible organizations, determining active shifts, retrieving qualified workers, evaluating task load and performance, applying organization rules, and assigning tasks using a balanced workload strategy with ties resolved on performance metrics.
Identifies the data used: open unassigned tasks, task zone and activity, organizations and inclusion or exclusion rules, active shifts and the department time zone from the schedule generation profile, workers on those shifts and their qualifications and competencies, current task load, average task completion time and completed task volume.
Describes automatic assignment at each scheduled run with no approval, review or acceptance step, and states that when worker capacity is reached remaining tasks stay unassigned until a later run rather than being routed to a human for decision.
Documents the agent template, the Task Assignment Rules prompt node with Include All Organizations, organization inclusion and exclusion lists, Maximum Number of Tasks Allowed per Worker, Assignment Window Before Shift Start, Assignment Window After Shift Start and Task Type Assignment Order, the Agent Team Settings schedule trigger, zone competencies, and the ORA_ASE_SAS_INTEGRATION_ENABLED profile option.
Lists the duty roles required to administer SCM AI agents and to interact with agents in product pages, the privileges required to interact with this agent, and the new data security instance sets for Workforce Schedule and Scheduling Workload Plan data, without stating the identity under which the scheduled agent run itself writes assignments.
States that the agent is delivered as a preconfigured AI Agent Studio template that must be used or copied to create an agent team and given a schedule trigger before it runs, and that a site-level profile option must be set for permission groups.
States that the feature requires Oracle HCM Workforce Scheduling to be set up with worker schedules generated, published and integrated with Inventory Management, and points to the 26A Redwood: Assign On-Shift Employees to Tasks feature for that integration.
Establishes the documented scope as assignment of existing open tasks to workers, with unassigned tasks carried to the next scheduled run, and contains no statement of task creation, cancellation, status change, priority change or execution monitoring by the agent.
Describes only retrieval of Oracle and Workforce Scheduling data and assignment of Oracle inventory tasks, with no documented command, message or dispatch to any external system.
- Inventory Task Allocation Assistant - Assign tasks to robotic material handling equipment (Inventory Management 26D What's New)
Supports: Actions · Availability · Admin controls · Permission basis · Primary source
States that the agent can automatically assign receipt delivery tasks to robotic material handling equipment when receipt deliveries are configured to be performed by robots, with delivery carts listed in the Task Assignment Rules node - an assignment target change within Oracle, documented for a forthcoming release.
Appears in the Inventory Management 26D What's New, a readiness document whose revision history records creation on 04 SEP 2026, so the robotic assignment behaviour is forthcoming-release documentation rather than established current functionality.
States that enabling this behaviour requires both the 25D robotic delivery feature setup and the 26B agent setup, plus configuration of the delivery cart list in the Task Assignment Rules prompt.
Repeats the same three user privileges required to interact with the agent, with each marked as an existing privilege rather than a new agent-specific role.
- AI Agent: Inventory Tasking Assistant (Inventory Management 26A What's New)
Supports: General · Limitations · Primary source
Describes the separate 26A Inventory Tasking Assistant, which recommends task-to-operator and operator-to-zone assignments, helps managers assign, unassign and reassign tasks, suggests zone deployment after layout changes and enables operator self-assignment, and is added to a Redwood page through a guided journey - establishing it as a distinct, human-invoked, recommendation-centred agent rather than an earlier name for the Task Allocation Assistant.
Provides the contrast used to keep this record scoped to the Task Allocation Assistant: the Tasking Assistant recommends and assists a human at the point of use, while the Task Allocation Assistant runs on a schedule and assigns automatically.
- Redwood: Deliver Goods with Robotic Material Handling Equipment (Inventory Management 25D What's New)
Supports: Limitations · External actions · Primary source
Establishes that physical robotic delivery is performed by an external partner robot system: material is picked onto a named delivery cart, the cart name in Oracle must match the cart name in the partner system, and when the autonomous delivery robot drops off the cart Oracle updates the delivery to Delivered with a Delivered by Autonomous Guided Robot note.
Shows that the Oracle-to-robot boundary is a separate deterministic integration feature with its own setup parameters, not an action taken by the AI agent.
- Supply Chain & Manufacturing features with AI (Oracle Fusion AI feature list)
Supports: Availability · General · Primary source
Lists AI Agent: Inventory Task Allocation Assistant under Advanced Inventory Management with release Update 26B and type AI agent, and lists the robotic equipment entry as Update 26D.
Lists AI Agent: Inventory Tasking Assistant (Update 26A) and AI Agent: Inventory Task Allocation Assistant (Update 26B) as separate features in the same table, supporting the decision to treat them as distinct capabilities.
- Inventory Management 26B Feature Summary
Supports: Default state · Licensing · Primary source
Lists the feature with Action to Enable Setup Required and Impact to Existing Processes None, where Oracle defines None as features that are not delivered enabled and will need steps to use, which the customer chooses if and when to enable.
Places the feature in the Advanced Inventory Management module of Oracle Fusion Cloud Inventory Management, the product surface whose setup and licensing information Oracle directs implementers to before enabling these features.