Customer Escalation Workup

Build an agent that researches a high-priority support ticket across Jira, your customer database, and Notion, then generates a thorough escalation brief with context, a timeline, and recommended actions.

The Business Problem

When a high-priority support ticket lands, your team scrambles to assemble context. Someone checks Jira for ticket history, another person digs through the customer database for account details, and a third searches Notion for related documentation. This takes 30-60 minutes per escalation, and critical details often get missed.

You need a repeatable process that pulls everything together in minutes, not hours.

What You'll Build

Component	Details
Agent	Support Researcher
Workflow	4-step escalation research workflow
Guidelines	3 rules for data handling, structure, and escalation triggers
Connections	Atlassian (Jira), PostgreSQL, Notion
Output	Escalation brief artifact

Step 1: Set Up Connections

Go to Connections in the sidebar and add the following:

Atlassian (Jira)

1. Click Add Connection and select Atlassian.
2. Click Connect and sign in with your Atlassian account.
3. Select your Atlassian site and authorize access.

PostgreSQL (Customer Database)

1. Click Add Connection and select PostgreSQL.
2. Enter your customer database connection string.
3. Click Connect.

Notion

1. Click Add Connection and select Notion.
2. Click Connect and authorize access to your Notion workspace.
3. Confirm the connection can see the pages and databases you need.

WARNING

This workflow handles sensitive customer data. Make sure all three connections use credentials with the minimum permissions needed — read-only access is ideal.

Step 2: Create the Agent

Go to Agents and click New Agent.

Field	Value
Name	Support Researcher
Description	Researches customer context across systems for support escalations
Preferred Model	Claude (default)
Risk Tolerance	Low
Plan Approval	Required

For the agent's Role Instructions, use:

You are a support researcher who assembles escalation briefs for the customer success team. You are thorough, concise, and careful with sensitive information. You never include raw personal data in reports. You always provide actionable recommendations.

INFO

Low risk tolerance means the agent takes a cautious approach — it prefers to ask for clarification rather than make assumptions. Plan approval required means you review the agent's plan before it executes, giving you a checkpoint before any data is pulled.

Step 3: Add Guidelines

Create three guidelines and assign them to the Support Researcher agent.

Guideline 1: No Raw PII

Field	Value
Name	Never include raw customer PII in reports
Content	Do not include raw email addresses, phone numbers, physical addresses, or payment information in any output. Reference customers by account name and ticket ID only. If specific contact details are needed, note "contact details available in the customer database" rather than including them directly.
Activation	Always active

Guideline 2: Recommended Actions

Field	Value
Name	Always include a recommended action section
Content	Every escalation brief must end with a "Recommended Actions" section containing 2-5 specific, prioritized next steps. Each action should name who should take it and by when. Frame actions as decisions, not tasks — e.g., "Decide whether to offer a contract extension" rather than "Send an email."
Activation	Always active

Guideline 3: High-Value Escalation

Field	Value
Name	Escalate to manager if contract value exceeds $100K
Content	If the customer's contract value (annual or total) exceeds $100,000, add a prominent notice at the top of the brief: "HIGH-VALUE ACCOUNT — Manager review required before response." Include the contract value and renewal date in the summary section.
Activation	Always active

Step 4: Build the Workflow

Go to Workflows and click New Workflow.

Field	Value
Name	Customer Escalation Workup
Description	Researches a flagged support ticket and generates an escalation brief
Default Agent	Support Researcher

TIP

This workflow is designed to be triggered manually when a ticket is escalated. You can also set up a hook to trigger it automatically when a Jira ticket is labeled "escalated."

Step 1: Pull Jira Ticket Details

Field	Value
Type	Action
Instructions	Look up the Jira ticket specified in the input. Retrieve the ticket summary, description, priority, status, assignee, reporter, creation date, and all comments. Also pull the last 5 tickets from the same reporter to identify patterns.
Auto-Approve	No

Step 2: Look Up Customer in Database

Field	Value
Type	Action
Instructions	Query the customer database for the account associated with this ticket. Retrieve the account name, contract value, renewal date, customer success manager, account tier, and any open opportunities. Query the support interactions table for the last 3 interactions logged for this account.
Auto-Approve	No

Step 3: Search Notion for Related Docs

Field	Value
Type	Action
Instructions	Search Notion for any documentation related to this customer or the issue described in the ticket. Look for meeting notes, onboarding documents, feature requests, and previous escalation briefs. Summarize anything relevant.
Auto-Approve	No

Step 4: Create Escalation Brief

Field	Value
Type	Action
Instructions	Compile all research into a structured escalation brief with these sections: (1) Executive Summary — 2-3 sentence overview, (2) Customer Profile — account details, contract info, relationship history, (3) Issue Timeline — chronological sequence of events, (4) Root Cause Analysis — what went wrong and why, (5) Impact Assessment — business impact to the customer, (6) Recommended Actions — prioritized next steps. Save as an artifact.
Auto-Approve	No

WARNING

All four steps have Auto-Approve turned off. Because this workflow handles sensitive customer data, you review each step before the agent proceeds. This adds a few minutes but ensures nothing inappropriate is pulled or included.

Step 5: Run the Workflow

1. Open the workflow and click Run.
2. When prompted, enter the Jira ticket ID (e.g., SUP-4521).
3. Review and approve each step in the Runs tab.
4. After Step 4, find the completed escalation brief in Artifacts.

Expected Output

The workflow creates one artifact:

Artifact: "Escalation Brief — [Ticket ID] — [Customer Name]"

A structured document like this:

HIGH-VALUE ACCOUNT — Manager review required before response.

Executive Summary Acme Corp (Annual Contract: $250K, renewing Aug 2026) reported a critical data sync failure affecting their production environment. The issue has been open for 72 hours across 3 Jira tickets. Previous escalation in November 2025 was resolved with a hotfix.

Customer Profile

• Account: Acme Corp | Tier: Enterprise
• CSM: Sarah Chen | Contract: $250K/yr | Renewal: Aug 2026
• Relationship: 2-year customer, generally satisfied, NPS 8

Issue Timeline

• Feb 27: Data sync failure reported (SUP-4519)
• Feb 28: Workaround attempted, partial success (SUP-4520)
• Mar 1: Full production impact, priority escalated (SUP-4521)

Root Cause Analysis API rate limiting change in v2.3 release caused sync jobs to fail silently...

Impact Assessment Customer cannot process daily transactions, estimated revenue impact of $15K/day...

Recommended Actions

1. Engineering: Deploy rate limit fix within 24 hours (owner: Dev Lead)
2. CSM: Schedule call with Acme Corp VP of Ops by end of day (owner: Sarah Chen)
3. Leadership: Decide whether to offer SLA credit for 3-day outage (owner: VP CS)

Monitoring and Cost

Each run of this workflow typically uses:

• Steps 1-3 (research): ~1,500 tokens each
• Step 4 (compilation): ~3,000 tokens

Because plan approval is required and all steps need manual approval, expect each run to take 5-10 minutes of your time. The agent does the heavy lifting — you just review and approve.

Troubleshooting

Problem	Solution
Jira ticket not found	Verify the ticket ID format matches your project key (e.g., SUP-1234). Check the Jira connection has access to the project.
Database returns no matching account	The agent searches by company name from Jira. If the names don't match across systems, add the account name to the Jira ticket before running.
Notion search returns too many results	Narrow the search by adding the customer name to the Notion step instructions.
Brief includes PII despite the guideline	Strengthen the guideline with specific field names to exclude, or add examples of acceptable vs. unacceptable references.