Skip to main content
Once you have agents, tools, and datasources set up, the next level is automation. Workflows let you chain multiple agents together, Jobs let you schedule execution, and Triggers let you respond to events automatically.

​
Workflows: Chaining Agents

A Workflow is a multi-step sequence that combines multiple agents, tools, and logic to solve complex problems.

​
What is a Workflow?

Instead of single agent handling everything, workflows let different agents specialize: Single agent (limited): Customer → Support Bot → Answer or Escalate Workflow (powerful):
  1. Customer Input
  2. Agent 1: Classifier — Is this a FAQ question? (Yes / No)
  3. Workflow splits:
    • Yes → FAQ Bot
    • No → Support Ticket Agent
  4. Response

​
Workflow Components

​
Common Workflow Patterns

Steps run one after another:Step 1 → Step 2 → Step 3Example: Extract info → Validate → Save to database
Route based on conditions:Check type?
  • If Sales → Route to Sales Agent
  • If Support → Route to Support Agent
  • If Admin → Route to Admin Agent
Run multiple steps simultaneously:Input runs in parallel across:
  • Agent A (processes data)
  • Agent B (checks quality)
  • Agent C (generates report)
Then combine results.
Handle failures gracefully:Try Step 1:
  • Success → Continue to Step 2
  • Failure →
    • Run Fallback Agent
    • Log error
    • Alert team

​
Creating a Workflow

1

Define Steps

List what agents/tools will do and in what order
2

Choose Agents

Select which agents participate in the workflow
3

Define Variables

Specify what data flows between steps (input, output)
4

Set Logic

Define routing, conditions, and error handling
5

Test

Run the workflow with test data
6

Deploy

Make it available for Jobs and Triggers

​
Jobs: Scheduled Execution

A Job runs a workflow or agent on a schedule.

​
Why Use Jobs?

  • Recurring Tasks - Run daily, weekly, monthly
  • Batch Processing - Process 1000s of records automatically
  • Maintenance - Clean up data, generate reports
  • Monitoring - Check status at regular intervals

​
Job Scheduling

Workflow: “Daily Sales Report”
  • Run: Every day at 2:00 AM
    • Process yesterday’s sales data
    • Generate report
    • Email to team
    • Log results
  • Next run: Tomorrow 2:00 AM
  • Next run: Day after 2:00 AM (continues indefinitely or until disabled)

​
Job Configuration

​
Schedule Patterns

  • Every day
  • Every Monday
  • Every 2 hours
  • Every 15 minutes

​
Job Status & Monitoring

Job: “Daily Report”
  • Last Run:
    • Status: Success ✅
    • Started: 2:00 AM
    • Duration: 3 min 42 sec
    • Records processed: 1,245
  • Next Run:
    • Scheduled: Tomorrow 2:00 AM
  • History:
    • Yesterday: Success (3m 15s)
    • 2 days ago: Success (3m 51s)
    • 3 days ago: Failed (timeout)

​
Triggers: Event-Based Automation

A Trigger automatically runs a workflow or agent when something happens.

​
Types of Triggers

Webhook

Triggered by HTTP POST to an endpoint. External systems can trigger your workflows.

Schedule

Same as Jobs - run on a schedule

Manual

User manually clicks a button to run

Event

Triggered by events in PLai (agent completion, data arrival, etc.)

​
Webhook Trigger Example

  1. External System sends an HTTP POST (new customer) to PLai Framework
  2. PLai Framework receives it via the Webhook Trigger
  3. Run Workflow: “Onboard Customer”
    • Agent: Send welcome email
    • Agent: Create account
    • Agent: Log to CRM
  4. Response sent back to External System: “Workflow started”

​
Trigger Configuration

​
Webhook Security

  • ✅ Unique URLs (hard to guess)
  • ✅ Rate limiting (prevent abuse)
  • ✅ Signature verification (validate requests)
  • ✅ Audit logging (track all calls)

​
Real-World Examples

​
Example 1: Lead Scoring Pipeline

Trigger: New form submission (webhook) Workflow: Lead Scoring:
  • Agent 1: Extract info (name, company, industry)
  • Agent 2: Score lead (1-10 based on data)
  • Agent 3: Route to sales rep
  • Tool: Send email to sales rep
  • Tool: Log to CRM
Result: Within seconds, lead scored and routed

​
Example 2: Nightly Data Processing

Job: Daily data ETL — Schedule: Every day at 2:00 AM Workflow:
  • Agent 1: Extract data from API
  • Agent 2: Transform/clean data
  • Agent 3: Validate quality
  • Tool: Load to data warehouse
  • Tool: Generate report and email
Result: Every morning, team has fresh data

​
Example 3: Multi-Step Support Process

Trigger: Customer submits support ticket (email or form) Workflow: Support Process:
  1. Agent 1: Classify issue
    • If FAQ → Route to FAQ Bot
    • If urgent → Route to support team
  2. If FAQ issue:
    • Agent 2: Answer from FAQ
  3. If urgent:
    • Agent 3: Create ticket
    • Tool: Send to support queue
    • Tool: Notify team
Result: Automated FAQs, urgent issues escalated immediately

​
Best Practices

Keep Workflows Simple

Start with 2-3 steps. Add complexity as needed. Complex workflows are harder to debug.

Test Before Scheduling

Test workflows manually before setting up automatic jobs or triggers

Monitor Execution

Check job/trigger logs regularly. Set up alerts for failures.

Document Variables

Clearly document what data flows between steps to prevent errors

​
Workflow Naming

✅ Good names:
  • “Lead Scoring Pipeline”
  • “Daily Sales Report”
  • “Customer Onboarding”
  • “Data Quality Check”
❌ Poor names:
  • “Workflow 1”
  • “Process”
  • “Automation”
  • “Run”

​
Monitoring & Debugging

​
Viewing Execution History

Workflow: “Lead Scoring”

​
Debugging Common Issues

Check:
  • Are agents doing unnecessary work?
  • Can steps run in parallel instead of sequentially?
  • Are tools timing out?
  • Is datasource search slow?
Check:
  • External APIs timing out?
  • Rate limits being hit?
  • Data format inconsistencies?
  • Credentials expiring?
Check:
  • Is job enabled?
  • Is timezone set correctly?
  • Is workflow available?
  • Any recent errors?
Check:
  • Is endpoint accessible?
  • Payload format correct?
  • Rate limit exceeded?
  • Check webhook logs for requests

​
Key Takeaways

✅ Workflows - Chain agents together for complex tasks
✅ Jobs - Schedule workflows to run automatically on a schedule
✅ Triggers - Run workflows in response to events (webhooks, schedules, etc.)
✅ Automation at Scale - Handle routine work without manual intervention

​
Next Steps