Team Settings

This guide explains the behavior settings available when configuring AI teams in Deeployd.

Collaboration Mode

Determines how agents in a team work together on tasks.

Sequential

Agents work one after another in a defined order. Each agent completes their work before passing to the next.

Task → Agent A → Agent B → Agent C → Result

Best for:

• Workflows with clear stages (lead qualification → demo scheduling → follow-up)
• When each stage depends on the previous stage's output
• Quality control checkpoints between stages

Parallel

All agents work simultaneously on different aspects of a task.

        ┌→ Agent A ─┐
Task ───┼→ Agent B ─┼→ Combine → Result
        └→ Agent C ─┘

Best for:

• Tasks that can be split into independent subtasks
• Research from multiple sources
• When speed is more important than strict ordering

Hybrid

Combination of sequential and parallel. Some agents work in parallel, then hand off to sequential stages.

        ┌→ Agent A ─┐
Task ───┤           ├→ Agent D → Agent E → Result
        └→ Agent B ─┘

Best for:

• Complex workflows with both independent and dependent work
• When some tasks need multiple agents working together, then a single agent synthesizing
• Enterprise processes with parallel research and sequential approval

Execution Mode

Controls how much autonomy the team has in executing tasks.

Mode	Description	Human Involvement
Manual	Every action requires human approval	Maximum control, slower
Plan First	Team creates a plan, waits for approval, then executes	Balance of control and speed
Supervised	Team executes but flags important decisions	Light oversight, faster
Automatic	Team executes autonomously, humans notified of outcomes	Minimal involvement, fastest

When to Use Each Mode

Manual

• Testing new agents or workflows
• High-stakes tasks with significant consequences
• Regulatory requirements demand human-in-the-loop

Plan First

• Complex tasks where you want to review strategy
• Building trust with a new team configuration
• Understanding agent reasoning before execution

Supervised

• Team has proven reliable on similar tasks
• Tasks have occasional edge cases needing human judgment
• You want to be informed without blocking execution

Automatic

• Team has high success rate on task type
• Tasks are routine and well-defined
• Speed is critical

Agent Autonomy Level

Defines how proactively agents can act within the team.

Reactive

Agents only act when given explicit tasks. They wait for instructions.

You: Qualify this lead
Agent: [Qualifies the lead]
Agent: Done. Waiting for next instruction.

Proactive

Agents can suggest actions and identify opportunities, but ask before acting.

Agent: I noticed 5 leads haven't been contacted in a week.
Agent: Should I send follow-up emails to them?
You: Yes, go ahead.
Agent: [Sends follow-ups]

Autonomous

Agents identify and execute tasks on their own within their defined scope.

Agent: [Monitors CRM continuously]
Agent: [Automatically contacts new leads matching ICP]
Agent: [Schedules follow-ups based on engagement]
Agent: [Escalates complex cases to humans]

Choosing Autonomy Levels

Level	Trust Required	Speed	Control
Reactive	Low	Slow	High
Proactive	Medium	Medium	Medium
Autonomous	High	Fast	Lower

A2A Visibility

Controls how much of agent-to-agent (A2A) communications users can see.

Hidden

No A2A conversations visible. Users only see final outcomes.

Use when:

• Users don't need to see internal processing
• Simplicity is more important than transparency
• Tasks are routine and trusted

Summary

Brief summaries of agent conversations. Shows what happened without full details.

Use when:

• Users want oversight without information overload
• Debugging needs to be possible but not constant
• Balance between transparency and usability

Full

Complete transcripts of all agent-to-agent messages. Maximum transparency.

Use when:

• Building trust in a new team
• Debugging complex multi-agent interactions
• Regulatory requirements demand full audit trail
• Training or improving agent behavior

Play/Pause (Safety Control)

The Play/Pause control is a safety mechanism for controlling team activity.

States

State	Description
Running	Team is active, processing tasks, agents communicating
Paused	All activity frozen, tasks queued but not processed
Stopped	Team is completely inactive

When to Pause

Use Pause when you need to:

• Stop runaway costs - If agents are in an unexpected loop
• Prevent unwanted actions - Before agents take irreversible actions
• Investigate issues - Freeze state to debug without ongoing changes
• Scheduled maintenance - Pause during system updates

What Happens When Paused

• All active agent processing stops immediately
• New tasks are queued, not dropped
• A2A conversations are frozen mid-thread
• When resumed, queued tasks are processed in order
• No data is lost during pause

Team First Start

When starting a team for the first time, agents go through an onboarding process:

1. Self-Discovery Phase

Agents analyze their own configuration:

• Review their instructions and capabilities
• Identify available tools
• Understand their role within the team

2. Tool Discovery Phase

Agents verify their tool access:

• Test connections to configured services
• Verify permissions are working
• Report any access issues

3. Handshake Phase

Agents introduce themselves to each other:

• Share their capabilities
• Establish communication protocols
• Define handoff procedures

4. V2MOM Alignment (Optional)

Team aligns on goals:

• Vision - What are we trying to achieve?
• Values - What principles guide us?
• Methods - How will we work together?
• Obstacles - What challenges might we face?
• Measures - How do we know we're succeeding?

Trigger Configuration

When deploying a team from a template, you can configure automation triggers during setup:

Event Triggers

Enable events from your connected integrations to automatically trigger agent actions:

• CRM Events: New leads, deal stage changes, tickets
• Communication Events: Slack mentions, messages
• Development Events: PRs, issues, deployments

Events are organized by integration. Only events from connected integrations can be enabled.

Scheduled Tasks

Configure recurring tasks that run automatically:

Schedule	Example
Morning Pipeline Review	Weekdays at 9:00 AM
Weekly Forecast	Mondays at 10:00 AM
SLA Compliance Check	Every 2 hours
Daily CSAT Review	Weekdays at 9:00 AM

Schedules use cron expressions for flexibility. Enable or disable suggested schedules based on your needs.

Bootstrap Tasks

One-time setup tasks that run when the team is first deployed:

• Import Pipeline - Analyze and categorize existing CRM data
• Setup Qualification Framework - Configure lead scoring criteria
• Create Meeting Prep Templates - Set up standard briefing documents
• Import Knowledge Base - Load documentation for support teams
• Configure Escalation Rules - Define escalation paths

Bootstrap tasks have priority levels (high, medium, low) and suggested due times.

Proactive vs Reactive Teams

The trigger configuration reflects the team's behavior pattern:

Pattern	Characteristics	Example Teams
Proactive	Self-generating work, schedules, outreach	Sales, Marketing
Reactive	Event-driven, responds to triggers	Support, DevOps
Hybrid	Both proactive and reactive	Customer Success

Proactive teams typically have more schedules and bootstrap tasks. Reactive teams rely more on event triggers.

Recommended Configurations

Conservative (High Control)

Collaboration: Sequential
Execution: Plan First
Autonomy: Reactive
A2A Visibility: Full

Best for: New deployments, sensitive data, regulated industries

Balanced

Collaboration: Hybrid
Execution: Supervised
Autonomy: Proactive
A2A Visibility: Summary

Best for: Established teams, moderate risk tasks

High Autonomy

Collaboration: Parallel
Execution: Automatic
Autonomy: Autonomous
A2A Visibility: Hidden

Best for: Proven teams, routine tasks, high-volume operations

Changing Settings

You can adjust team settings anytime through:

Dashboard

1. Go to Teams page
2. Click Edit on your team
3. Navigate to Team Behavior tab
4. Adjust settings and save

Loyd

Simply ask Loyd:

• "Change my sales team to automatic execution mode"
• "Make my support team more autonomous"
• "Show me full A2A visibility for the DevOps team"

API

await client.teams.update('team-123', {
  orchestration: {
    executionMode: 'automatic',
    autonomyLevel: 'autonomous',
    a2aVisibility: 'summary'
  }
})

note

Changing settings while a team is running applies to new tasks only. Existing tasks continue with their original settings.

---

Next: Learn about Agent-to-Agent Communication for more details on how agents collaborate.