Support Assistant

Goal: Answer support questions from your docs — consistently, with citations — and draft replies in minutes, not hours.

Architecture Overview

A practical support stack in Springbase has four layers:

Support Knowledge Context  →  Recipes (triage + reply)  →  Chat (daily use)  →  Agents (optional actions)
         ↑
    Live Data (help center)

You do not need Agents on day one. Start with Context + Chat + one Reply Recipe.

Week 1 — Build the Knowledge Base

Step 1: Create the Context

1. Contexts → Create New Context
2. Name: Support Knowledge Base (or your product name + Support)
3. Add description: "Official help docs, policies, and resolved ticket patterns."

Step 2: Add trusted sources

Upload static material first:

Remove duplicates, outdated drafts, and anything you would not want cited to customers.

Step 3: Add Live Data (Pro/Max)

On the Live Data tab:

• Add your help center sitemap or RSS (daily or weekly refresh)
• Optionally sync a Google Drive folder where docs team publishes updates

See Live Sources for setup details.

Step 4: Test grounding

In Chat, attach the Context and ask hard questions only your docs answer:

@Support-Knowledge-Base What is our refund policy for annual plans cancelled within 30 days?

Require citations. If answers are generic, tighten the question or clean noisy documents from the Context.

Week 2 — Create Support Recipes

Recipe 1: Draft customer reply

Prompt pattern:

You are a support specialist. Using ONLY the attached Support Knowledge Base,
draft a reply to {{customer_name}} ({{tier}} customer).
Customer message: {{ticket_body}}
Tone: {{tone}}
Include: direct answer, next steps, link to relevant doc section if applicable.
If the KB does not contain enough information, say what is missing — do not invent policy.

Attach the Support Context in Recipe settings (always-on grounding).

Recipe 2: Triage and classify

Variables: ticket_body, product_area (select)
Output format: JSON or table with Category, Severity, Suggested owner, One-line summary

Use before the reply Recipe on complex tickets.

Recipe 3: Escalation brief

For P1/P2: summarize impact, customer tier, repro steps, and recommended escalation path — grounded in SOP documents in the Context.

Daily Workflow

For ad-hoc policy questions, skip Recipes and use Chat with @Support-Knowledge-Base.

Optional: Agents for Handoffs

If your team uses Slack for escalations:

1. Connect Slack in Tools Marketplace.
2. Create an Agent Recipe: "Post escalation summary to #support-escalations" with variables for ticket ID and summary.
3. Always include ask before posting in the prompt.

Only document apps your team actually connected.

Optional: Meeting Automations

When support runs customer calls:

1. Capture meeting → Save to Context (Support or Account-specific Context).
2. Post-meeting automation runs a recap Recipe for the CSM handoff note.

See Meetings and Sales Follow-Up for the automation pattern.

Team Sharing

Read Security and Sharing before publishing anything public or unlisted with customer data patterns.

Maintenance Rhythm

Generate a Custom Plan

Visit Help in the app, choose role Support, and run the playbook generator for a tailored checklist based on your workspace.

Related Guides

• Contexts — structure and permissions
• Recipes — variables and scheduling
• Troubleshooting — generic answers, noisy Contexts