8.1 KiB
Client Lifecycle
Overview
The client lifecycle defines the end-to-end journey from first contact through ongoing engagement, renewal, and offboarding. Every stage preserves the reconnaissance-versus-execution rule and evidence-tier discipline.
Reconnaissance-versus-Execution Rule: Agents may detect and draft. Humans approve before publish. This rule applies at every stage where an agent proposes a client-facing change.
Evidence-Tier Rule: Verified Evidence (concrete, timestamped, observable) may be presented as confirmed issues. Indicative Evidence (sampled, directional, probabilistic) must be presented as directional risk or opportunity, not confirmed customer loss. Ambiguous findings default to Indicative Evidence unless upgraded by human review. Rejected Verified Evidence classifications default to Indicative Evidence unless the reviewer discards the finding or requests additional evidence.
Stage 1: Prospect
Initial identification of a potential client. No engagement yet. May come from referral, inbound, outbound, or partner channel.
Activities:
- Fit assessment against ideal client profile
- Preliminary digital presence scan (public signals only)
- Decision: qualify or disqualify
Output: Qualified prospect record or disqualification note.
Stage 2: Initial Assessment
Structured diagnostic of the prospect's digital pathways affecting discovery, trust, and customer action.
Activities:
- AI visibility assessment (search + AI-assisted discovery)
- Customer Path Integrity review (GBP, website, booking, contact, citations)
- Local entity consistency check
- Competitor baseline snapshot
- Evidence collection following evidence-tier rules
Output: Initial Assessment Report — findings categorized by evidence tier, prioritized by impact and fixability.
Gate: Human review of assessment before any client-facing delivery.
Stage 3: Proof-of-Concept Fix
During the proof-of-concept stage, agents may detect issues, gather evidence, and draft proposed fixes.
Reconnaissance-versus-Execution Rule: Agents may detect and draft. Humans approve before publish. This rule applies to every proposed client-facing change.
Proof-of-Concept Fix Approval Gate: Any fix that publishes to a client-facing surface requires human approval before it goes live. This includes, but is not limited to:
- Google Business Profile edits
- Review responses
- Q&A answers
- Business descriptions
- Service descriptions
- Hours
- Phone numbers
- Booking links
- Website changes
- Directory or citation updates
N≥100 Zero-Error Graduation Threshold: For any specific task type (e.g., GBP description edits, review responses, Q&A answers), the agent must demonstrate ≥100 successful executions with zero errors before that task type is eligible for reduced-review or auto-approval workflows. The threshold is task-type-specific, not aggregate.
Evidence Collection: Each proposed fix must cite its evidence tier. Verified Evidence fixes (wrong phone, broken link, incorrect hours) proceed with clear confirmation. Indicative Evidence fixes (AI answer absence, weak entity signals) are labeled as directional.
Output: Approved fixes applied; Before/After evidence captured.
Stage 4: Before/After Report
Documentation of what was found, what was fixed, what remains, and the evidence tier for each item.
Structure:
- Verified Evidence fixes — confirmed before/after with artifacts
- Indicative Evidence improvements — directional comparison, labeled as sampled
- Remaining risks — categorized by evidence tier
- Recommended next steps
Gate: Human review and approval of report before client delivery.
Stage 5: Retainer Offer
Formal proposal for ongoing engagement based on proven value from the proof-of-concept.
Components:
- Scope definition (monitoring domains, cadence, deliverables)
- Evidence-tier reporting standards
- Approval-gate framework for ongoing optimizations
- Pricing tied to recoverable LTV (20–30% ceiling per policy)
- Term and renewal terms
Gate: Human approval of offer terms before presentation.
Stage 6: Ongoing Monitoring
Continuous digital oversight per the agreed scope.
Reconnaissance-versus-Execution Rule (restated for Ongoing Monitoring): Agents may detect and draft. Humans approve before publish. No agent-initiated client-facing publishes without human approval.
Monitoring Domains (per scope):
- AI visibility (search + AI-assisted discovery sampling)
- Customer Path Integrity (GBP, website, booking, contact, citations)
- Local entity consistency
- Competitor signal tracking
- Review signal monitoring
- Structured data / schema health
Approval-Gate Framework:
- Verified Evidence findings → proposed fix drafted → human approval → publish
- Indicative Evidence findings → risk/opportunity noted → human review → optional fix drafted → human approval → publish
- N≥100 zero-error graduation threshold applies per task type before any reduced-review pathway
Cadence: Continuous monitoring with periodic human review checkpoints (minimum weekly).
Output: Monitoring alerts, drafted fixes pending approval, periodic summary.
Stage 7: Monthly/Periodic Reporting
Structured delivery of monitoring outcomes, actions taken, and forward look.
Report Contents:
- Verified Evidence fixes applied (confirmed before/after)
- Indicative Evidence risks/opportunities identified (directional, labeled)
- Monitoring alerts generated and dispositioned
- Competitor signal changes
- Evidence-tier compliance summary
- Recommended next actions with priority
Gate: Human review and approval of report before client delivery.
Stage 8: Renewal/Expansion
Evaluation and negotiation of continued or expanded engagement.
Activities:
- Value delivered review (verified fixes, risk reduction, visibility gains)
- Scope adjustment (add/remove monitoring domains, change cadence)
- Pricing re-evaluation against recoverable LTV
- Term extension or expansion agreement
Gate: Human approval of renewal/expansion terms.
Stage 9: Offboarding
Formal and complete disengagement when the relationship ends.
Required Actions:
- Final report delivery (status of all monitored pathways, remaining risks)
- Revocation of all agent credentials for client systems
- Removal of delegate access (GBP, social, directories, analytics, etc.)
- Revocation of API tokens, webhook subscriptions, integration keys where applicable
- Removal of client from monitoring systems, alerting, and data pipelines
- Confirmation that no automated agents retain access to client-facing surfaces
- Data retention / deletion per agreement and policy
Gate: Human confirmation that all access is revoked and client is fully removed from operational systems.
Output: Offboarding completion certificate signed by human operator.
Lifecycle Summary
| Stage | Name | Key Gate |
|---|---|---|
| 1 | Prospect | Fit qualification |
| 2 | Initial Assessment | Human review of assessment |
| 3 | Proof-of-Concept Fix | Human approval per fix (N≥100 threshold for task-type graduation) |
| 4 | Before/After Report | Human review of report |
| 5 | Retainer Offer | Human approval of terms |
| 6 | Ongoing Monitoring | Human approval per publish (reconnaissance-vs-execution) |
| 7 | Monthly/Periodic Reporting | Human review of report |
| 8 | Renewal/Expansion | Human approval of terms |
| 9 | Offboarding | Human confirmation of full access revocation |
Core Principles (Restated)
- Reconnaissance-versus-Execution: Agents detect and draft. Humans approve before publish. No exceptions.
- Evidence-Tier Discipline: Verified Evidence = confirmed issues. Indicative Evidence = directional risk. Ambiguous = Indicative by default. Rejected Verified = Indicative unless discarded or re-evidenced.
- N≥100 Zero-Error Graduation: Per task type, ≥100 successful executions with zero errors before reduced-review eligibility.
- Human-in-the-Loop: Every client-facing publish, every client-facing report, every scope/pricing decision requires human approval.
- Complete Offboarding: All access revoked, all tokens removed, all monitoring stopped, confirmed by human.