Jobs
Engagement Layer / Jobs
Intro
A job is an execution framework for coordinating operational tasks carried out by agents, automations, and enterprise systems to achieve specific business outcomes. Unlike workflows that follow predefined steps or directly perform actions, a job captures the intent, scope, policies, and constraints guiding how the tasks should be carried out. By setting the objective once, business users delegate the orchestration of required capabilities to Joule, monitoring the situation, making decisions within established policies, and engaging users when necessary. The value of jobs lies not only in automation itself but in reliably operationalizing automation and ensuring work remains observable and auditable.
Core Principles
Anatomy
Job Contract (Core)
Defines how the job operates and what outcome it ensures.
Includes:
- Goal (desired outcome)
- Scope (entities, data, timeframe)
- Triggers (event, schedule, manual)
- Policies/guardrails
- Autonomy level
- Allowed capabilities
- Ownership
- Escalation behavior
- Success criteria
Job contract
Job State
Represents the current lifecycle status of the job.
Typical states:
- Proposed
- Active
- Needs Attention
- Resolving
- Completed
- Closed
States change based on signals, execution progress, and user interaction.
Signals & Situations
Job signals
Jobs ingest and interpret system signals such as:
- Events
- Alerts
- Threshold breaches
- Tasks
These are:
- Correlated
- Filtered
- Aggregated into meaningful situations
Users never deal with raw system noise – only contextualized operational insights.
Job Runs
Job runs
A job is the persistent container.
A job run is a specific execution instance triggered by:
- User action
- Schedule
- Event
Example:
- Job: “Monitor margin anomalies”
- Runs: weekly checks, event-triggered detections
Job Cards
Job card
Primary representation of Jobs across all surfaces.
Contain:
- Job name + objective
- Current status
- “What’s happening now”
- Key signals or anomalies
- Latest outcome or required action
- Urgency indicator
Variants:
- Compact (lists, notifications)
- Expanded (spaces, conversations)
- Focused (needs attention)
Cards act as:
- Entry point
- Action surface
- Explanation layer
Job Detail View
Job detail view
Full inspection layer for:
- Execution contract
- History of runs
- Signals and outcomes
- Decisions and audit trail
- Capability usage
Used for:
- Oversight
- Debugging
- Governance
Job Types
Ad-Hoc Jobs
One-time execution for a specific outcome.
Manual Reusable Jobs
Saved patterns that can be triggered repeatedly.
Standing Jobs
Persistent jobs that remain active over time and supervise an ongoing operational concern.
Characteristics:
- Continuously active
- Event-driven, scheduled, or both
- Monitor signals over time
Autonomous Jobs
Standing or event-driven jobs that are allowed to take bounded action without user input.
Characteristics:
- Policy-bound autonomy
- Auto-resolve low-risk cases
- Escalate ambiguous, material, or policy-sensitive cases
- Require clear execution contract and guardrails
Behavior and Interaction
Job Creation
Jobs are created through:
1. Conversational creation (primary)
User expresses intent → Joule proposes job → user reviews and activates
2. Template-based creation
Predefined jobs configured via parameters
3. System-suggested jobs
System detects patterns → suggests creating a job
Qualification
Not every action becomes a job.
A job is created only if:
- Work persists over time
- Coordination is required
- A decision may be needed
- Accountability is required
Otherwise, interaction remains ephemeral.
Configuration
Users can:
- Adjust scope and thresholds
- Modify policies and autonomy
- Reassign ownership
- Extend capabilities (via system)
Changes are tracked as part of the Job history.
Surfaces
Jobs do not rely on a fixed UI.
Instead, they surface across interaction points.
Users experience jobs through a combination of:
1. Contextual surfaces (primary)
-
Conversations
- Job proposals
- Explanations
- Outcome summaries
-
Spaces
- Investigation
- Collaboration
- Decision-making
-
Notifications
- Alerts
- Escalations
- Required actions
2. Management surfaces (secondary)
-
Job list
- Overview of all jobs
- Filtering by status, ownership, recency
-
Job detail view
- Full execution context
Conversational Interaction Patterns
Conversations are the entry and explanation layer.
Users can:
- Create Jobs
- Inspect Jobs
- Ask for explanations (“why?”, “what changed?”)
- Adjust parameters
- Resolve decisions
Jobs appear as:
- Proposals
- Active status cards
- Outcome summaries
Explainability
Explainability is built into jobs at multiple levels:
During Creation
- Intent rephrasing
- Clarification questions
- Contract preview
During Execution
- “What is happening now?”
- Why are actions taken
- What signals triggered behavior
During Decision Moments
- Context
- Recommendation
- Impact of each option
Post Execution
- Outcome summaries
- Decision trace
- Capability usage
Explainability example
Governance
Ownership
Each job has an owner :
- Individual user
- Team
- Role
- Organization
Owners are responsible for:
- Oversight
- Decisions
- Resolution
Autonomy Levels
- Observe
- Recommend
- Assist
- Autonomous
Audit & Compliance
The system records:
- Job configuration
- Decisions
- Actions taken
- Capabilities used
- Complete execution history
The system ensures:
- Traceability
- Compliance
- Accountability