Virtual CX Team
The behind-the-scenes infrastructure team. Not here to respond to customers — here to make sure every system, template, and piece of knowledge is always in place so that AI agents doing the frontline work can perform at their best.
Build the Support Infrastructure, Not the Support
The Virtual CX Team operates behind the scenes. Their job is to build and maintain the systems, knowledge, templates, and feedback loops that frontline AI agents rely on. When those agents respond to customers, they're drawing on the work this team does every day.
🧠 Memory
- Shared memory pool: all agents read from a central knowledge base
- Per-agent private memory for work-in-progress and drafts
- Every update is timestamped and version-controlled
- Stale content flagged automatically after configurable TTL
- Feedback from frontline agents flows back into memory
- Manager has full read/write; specialists have scoped write access
🔐 Security
- No PII stored in team knowledge base — customer data stays at frontline layer
- All external publishing requires Manager sign-off
- Agents cannot modify shared memory without version control
- Audit log for all knowledge edits and template changes
- Access levels: Owner (Liis) → Manager → Agents
- New skills or capabilities require Owner approval before activation
🔄 Continuity
- Agent state persists across sessions — no cold starts
- Handoffs between agents include full context and task history
- Daily health check: each agent self-reports status to Manager
- Fallback routing if an agent is unavailable
- All work stored in structured output folders, accessible any time
- Weekly team sync: Manager compiles cross-agent learnings
⚡ Skills
- Skills are scoped per agent — no capability bleed between roles
- Skills version-controlled and documented
- Skills can be toggled on/off without restarting agent
- New skills go through: propose → review → test → approve → activate
- Owner can override any skill at any time
- Skills feed forward: Knowledge output feeds Copy, Copy output feeds dashboards
The Team
Four roles, one mission. The Manager coordinates and owns quality. The three specialists focus on their domain and feed their outputs back to the whole team.
CX Manager
Coordinates the team, owns the roadmap, routes work to specialists, ensures quality before anything leaves the team, and maintains the relationship with Liis.
🗺️ Responsibilities
Every new signal (feedback, request, product change, insight) comes here first. Manager decides urgency, which agent handles it, and what the output should look like.
Before anything produced by a specialist agent is used by the broader team or shared externally, Manager reviews it for accuracy, tone, and completeness.
Ensures Knowledge, Copywriter, and Analyst are aligned. Surfaces when one agent's output should inform another's work.
Gathers signals from frontline agents and routes them to the right specialist. Makes sure learnings don't get lost.
Summarises what the team produced, what was learned, what needs attention, and what's coming next.
🚦 Decision Rules
Route standard tasks to specialists, compile weekly reports, flag stale knowledge
Any output going externally, structural changes to team workflow, priority conflicts between agents
New agent creation, changes to guardrails, modifications to tone of voice guide
📋 Current Team Priorities Draft — to be refined with Liis
Give the Manager something
Strategy doc, brief, feedback summary, priority list
📋 What this agent does
Upload any document — product spec, policy update, FAQ, external article, past ticket export. The agent reads, structures, and indexes it into the KB.
Periodically checks existing KB entries against new information. Flags anything that may be outdated. Suggests updates rather than overwriting.
Can synthesise KB content into structured training docs, onboarding guides, or quick-reference sheets for frontline agents.
Other agents (and Liis) can query the KB directly — "What's the current refund policy?" — and get sourced, accurate answers.
📂 Knowledge Base ✓ 23 documents loaded · 272KB
🗺️ Help Center Sitemap
Upload to Knowledge Base
Product docs, policy files, FAQs, research, ticket exports, training materials
📋 What this agent does
Given a context or scenario, produces copy calibrated to the right platform — a Discord reply reads differently from a LinkedIn comment or a Reddit thread response.
Builds and curates a reusable library of response templates for common situations — shipping delays, feature requests, billing questions, compliments, complaints.
Reviews any copy before it leaves the team. Flags deviations from the Genspark voice. Can suggest rewrites that preserve the message but match the brand.
When the Knowledge Agent surfaces new product info or the Analyst flags a trend, this agent updates relevant templates proactively.
✋ Templates — Needs Your Approval
These 3 drafts are ready. Review each one, then approve or mark for revision. Nothing goes live until you say so.
🏷️ Platform Rules Draft — to refine
Discord — Casual, fast, human. Short sentences. Emojis OK. Community feel. Never corporate.
Instagram — Warm, visual, aspirational. Focus on emotion and experience. Minimal jargon.
X / Twitter — Direct, punchy, confident. Wit welcome. Never defensive. Get to the point.
Reddit — Honest, detailed, no marketing fluff. Reddit readers are smart and will call out anything fake.
Email — Clear, structured, personal. Subject line matters. No walls of text. Action-oriented.
LinkedIn — Professional but not stuffy. Thoughtful, on-trend, credibility-building.
Upload Brand Materials
Tone of voice guide, brand guidelines, existing templates, style references
📋 Provisional Scope ⚠ To be more clearly defined — placeholder
Keep agreed dashboards up to date as new data arrives. Flag when data is missing or stale.
Process customer feedback from all channels into structured themes. Surface what's recurring, what's new, what's getting better or worse.
Produce regular reports for the Manager. Highlight anomalies, shifts, and opportunities the team should act on.
When patterns suggest the KB is missing something or templates are performing poorly, flag it to the right agent.
⚠️ Needs Your Input
- · What dashboards need to exist?
- · What data sources will this agent have access to?
- · What decisions should this data inform?
- · Who else outside this team receives the reports?
Upload Data or Reports
Feedback exports, analytics CSVs, existing dashboards, reporting requirements