---
title: Support assistant
summary: >-
  A customer-facing support workflow that explains issues, next steps, and safe
  escalation paths.
agent_summary: >
  This workflow describes a support assistant that helps users resolve product,
  billing, and account issues while balancing trust, privacy, speed, and
  escalation.
canonical_url: /docs/workflows/support-assistant
page_type: workflow
related_resources:
  - /resources/workflows/support-assistant.v1.json
related_schemas:
  - /schemas/workflow.schema.json
  - /schemas/decision-record.schema.json
  - /schemas/verdict.schema.json
last_reviewed: '2026-04-09'
---
# Support assistant

A customer-facing support workflow that explains issues, next steps, and safe escalation paths.

> Agent summary: This workflow describes a support assistant that helps users resolve product, billing, and account issues while balancing trust, privacy, speed, and escalation.


## Headings
- ## Why this workflow matters
- ## What decisions exist inside the workflow
- ## Which guardrails apply
- ## Inputs and evidence required
- ## Example in practice
- ## Common drift patterns
- ## Escalation points
- ## Related resources and schemas
- ## Related pages

## Why this workflow matters

Support is one of the fastest places for trust to erode. A technically plausible answer that is too vague, too forceful, too slow, or too revealing can turn a solvable issue into a trust issue.

## What decisions exist inside the workflow

- what information to include or omit
- how confident to sound
- whether to escalate or self-serve
- how to structure next steps
- how much user context to use
- how fast and cheaply the system should respond

## Which guardrails apply

- `guardrail.brand-tone`
- `guardrail.safety-privacy`
- `guardrail.runtime-cost`
- `guardrail.provenance-escalation`

## Inputs and evidence required

- user request
- account or case context, if permitted
- consent and data classification state
- workflow surface
- locale
- traceable actor and toolchain metadata

## Example in practice

The two launch examples show the most common support failure modes:

- language that becomes pushy or unsupported even when the intent is helpful
- account workflows that need a secure handoff instead of a heroic assistant answer

## Common drift patterns

1. Tone becomes pushy or overconfident.
2. Sensitive data appears in the response.
3. The workflow hides uncertainty.
4. Runtime loops or retries degrade the experience.
5. Escalation is skipped even when the system lacks confidence.

## Escalation points

Escalate when the response involves sensitive account action, the system cannot verify the claim, a boundary is hit, or the user needs a human to intervene.

## Related resources and schemas

- Resource: `/resources/workflows/support-assistant.v1.json`
- Schema: `/schemas/workflow.schema.json`
- Decision schema: `/schemas/decision-record.schema.json`
- Verdict schema: `/schemas/verdict.schema.json`

## Related pages

- /docs/guardrails/brand-and-tone
- /docs/guardrails/safety-and-privacy
- /docs/examples/brand-tone-support-coercion
- /docs/examples/privacy-escalation

## Related pages
- /docs/guardrails/brand-and-tone
- /docs/guardrails/safety-and-privacy
- /docs/guardrails/runtime-and-cost
- /docs/examples/brand-tone-support-coercion
- /docs/examples/privacy-escalation

## Related resources
- /resources/workflows/support-assistant.v1.json

## Related schemas
- /schemas/workflow.schema.json
- /schemas/decision-record.schema.json
- /schemas/verdict.schema.json