A2CUSTOMER COMMS

Tier-2 support copilot

Sits inside the support agent's interface and drafts replies for complex or research-heavy tickets. Reads the incoming ticket, searches past resolved tickets and internal documentation for precedent, and produces a draft response the human agent edits and sends. Never sends autonomously. The pattern accelerates skilled agents on hard cases rather than handling easy cases without them, which is why it shows up in firms where the support workload is complex by nature and full deflection isn't realistic.

WHERE THIS FITS

BUSINESS SHAPES

B2B servicesProduct companyProfessional services

VOLUME THRESHOLD

Below 15 tier 2 tickets per agent per week a month, the payback rarely earns the build. Patterns this shape reliably pay back at 40+.

FITS BEST

Professional services, B2B SaaS with complex products.

PAYBACK · 3-5 moBUILD · Low-MediumVALUE · $35k-$120kWHEN · >5 support agents

FAILURE MODE TO DESIGN AROUND

Agents copy-paste without editing → robotic replies. Track edit rate as quality metric.

REQUIREMENTS · 4 REQUIRED, 1 OPTIONAL

Requirements describe capabilities the pattern needs in your environment, not the vendors you must buy. Any system that fills a requirement satisfies it — that’s what makes the catalog portable across the long tail of SMB tooling.

active_ticket_view
REQUIREDREAD + WRITErequest
The agent's working surface where they handle one ticket at a time. The pattern reads the open ticket and presents the draft inline.
DATA SHAPE
Currently-open ticket: full thread history, customer identifier, status, tags, and any internal notes. Pattern writes a draft into the reply composer or a side panel.
COMMONLY FILLED BY
- agent view inside a helpdesk system
- support interface embedded in a CRM
- custom internal support tool
ticket_history_corpus
REQUIREDREADcorpus
Previously resolved tickets the pattern searches for precedent. The source of the 'we've seen this before' value.
DATA SHAPE
Closed tickets with original inquiry, full response thread, resolution code, and time-to-resolve. PII stripped or vaulted before indexing.
COMMONLY FILLED BY
- archive of closed tickets in the helpdesk
- export of historical support cases
- knowledge management system fed by past resolutions
product_documentation_corpus
REQUIREDREADcorpus
Authoritative technical and procedural content the pattern grounds factual claims on.
DATA SHAPE
Technical documentation, runbooks, troubleshooting guides, configuration references. Often more structured than the customer-facing help center.
COMMONLY FILLED BY
- internal engineering wiki
- developer documentation site
- operations runbook collection
- shared drive of procedure documents
agent_feedback_signal
REQUIREDWRITEevent
Capture how much the agent edited the draft before sending. The single most useful quality signal this pattern produces.
DATA SHAPE
Original draft, final sent message, edit-distance score, optional agent rating thumbs-up/down. Per-ticket.
COMMONLY FILLED BY
- logged inside the ticket view as an internal note
- captured by an instrumented submit-handler
- lightweight thumbs widget next to the draft
customer_context_store
RECOMMENDEDREADrequest
Lookup of the customer's account, plan, history, and any prior tickets so the draft can reference the specifics.
DATA SHAPE
Customer records with account state, plan, recent activity, and prior ticket links.
IF MISSING
Drafts stay correct but generic. The agent has to fill in customer-specifics manually, which is most of the value being eroded.
COMMONLY FILLED BY
- CRM with customer profiles
- product database keyed by account
- billing system with subscription state

RUNTIME FLOW · 7 STEPS

01
Detect that the agent has opened a ticket in the working interface and read its content
active_ticket_view
02
Fetch customer context if available to enrich the draft with account-specific details
customer_context_store
DECISION Skip if customer_context_store not filled.
03
Retrieve similar past tickets and their resolutions via semantic search over ticket history
ticket_history_corpus
04
Retrieve relevant procedural and technical documentation
product_documentation_corpus
05
Generate a draft response with inline citations linking back to source ticket IDs and doc sections
06
Present the draft to the agent in the working interface, with citations and any flagged uncertainties
active_ticket_view
07
Capture the eventual sent message, compute edit-distance from the original draft, log the agent's feedback
agent_feedback_signal

EMISSIONS · 2

Structured outputs this pattern produces. Other patterns and client systems can subscribe to them, which is how the catalog composes over time.

draft_quality_signal
Per-ticket edit distance and optional rating, used to track whether the pattern is genuinely helping or producing drafts that always need rewriting.
CONSUMED BY
- quality dashboards
- prompt refinement workflows
- training data for fine-tuning if ever pursued
knowledge_gap_signal
Tickets where the pattern couldn't find good precedent or documentation, surfaced as gaps the team should fill in docs.
CONSUMED BY
- documentation maintenance workflows
- support team weekly retros