Compliance Field Copilot

4.1 Smart Template Recommender

Problem solved: Assessors waste time searching manually for the correct template (“Which one fits this site?”).

How it works: On site arrival, the copilot uses GPS location, facility type, assessment history, and organisation template library to rank the top 1–3 templates with confidence scores.

Example: “Based on your location at Warehouse B and your last 3 visits, I recommend the Fire Safety — Industrial template (92% match). Use this one?”

PLG link: Reduces time-to-value; improves single-session completion path.

4.2 Context-Aware Autofill

Problem solved: Repetitive manual entry of site details, assessor info, and recurring metadata.

How it works: Pre-populates fields from prior assessments at the same site, organisation defaults, and device context. Assessor confirms before save — never silent overwrite.

PLG link: Compresses target time-to-value from ~62 min toward ~28 min post-redesign baseline.

4.3 Voice-to-Form Capture

Problem solved: Two-step workaround — handwritten notes on-site, re-entered later at desktop.

How it works: Assessor speaks observations; speech-to-text + entity extraction maps content to structured form fields (location, issue type, severity, notes, evidence links).

Example: “Fire exit blocked at loading bay B, severity high.” → Copilot proposes field values; assessor taps Confirm.

Offline behaviour: Voice captured locally; transcription queues until connectivity returns or on-device model available.

PLG link: Directly attacks transcription-error and double-entry abandonment drivers.

4.4 Guided Completion & Criticality Alerts

Problem solved: Long questionnaires overwhelm field users; critical items get missed.

How it works: Copilot surfaces next required critical items, explains why they matter, and flags incomplete sections before submit. Severity-aware nudges for high-risk gaps.

Example: “3 critical items remain — including photo evidence for Section 4. This is required before submission.”

PLG link: Improves draft→sync→submit path (target growth from 5% to 38% of completions).

4.5 Pre-Submit Validation Agent

Problem solved: Fear that data will not sync correctly; submissions rejected for completeness errors.

How it works: Before submit, an agent runs validation rules: required fields, evidence attachments, consistency checks vs prior visits, policy thresholds. Returns pass/fail with fix list.

Guardrail: Copilot recommends fixes; assessor retains submit authority. High-severity compliance gaps escalate to manager queue.

4.6 PLG Propensity Intelligence (Admin Console)

Problem solved: Reactive discovery of activation failure and churn risk.

How it works: ML models trained on PLG funnel data predict:

• Activation risk — user unlikely to complete first assessment within 14 days
• Abandonment risk — mid-assessment dropout probability during offline/sync stages
• Churn risk — account with high licence assignment but low completion output
• Expansion PQL — account matching “3+ iPad assessments in 14 days incl. offline path” pattern

Outputs: Manager dashboard alerts, in-app nudges (“You have 1 draft ready to sync”), CS playbook triggers.

4.7 Manager Insights Copilot (Natural Language Analytics)

Problem solved: Compliance managers need answers without building custom reports.

How it works: LLM-RAG over assessment data, PLG metrics, and account metadata. Managers ask plain-language questions; copilot returns concise answers with drill-down links.

Example queries: “Which sites had the lowest completion rate this month?” · “Which accounts show expansion potential?” · “Where are assessors abandoning during offline sync?”