Canvassing AI
Use record-grounded field and manager assistance while keeping pricing, customer data, and record changes under human control.
Canvassing AI
Canvassing AI is an optional, connected assistant for short field guidance and manager analysis. It reads only records allowed by the signed-in organization and assignment scope. It never performs a mutation, sends a message, changes pricing, or saves extracted customer information.
The user needs AI View access, the organization must have AI enabled and configured, and usage must remain within the organization's credit and daily request limits.
Field Modes
Open a property and select Pre-knock or Objection, or open AI field coach to switch between the modes available for the current property and shift.
| Mode | What it does |
|---|---|
| Pre-knock | Builds a short opener from the supplied campaign and property-category facts |
| Objection | Suggests a calm response, one question, and a next step for a selected objection |
| Voice note | Structures a typed, pasted, or keyboard-dictated conversation transcript for review |
| Estimate | Identifies missing inputs and safe next steps from known visit or estimate context |
| Conversation | Gives concise coaching from the latest saved visit |
| Shift review | Summarizes aggregate shift performance and suggests one next improvement |
Objection choices include not interested, too expensive, talk to spouse, already has a company, just cleaned, leave a card, think about it, no time, renting, damage concern, plant concern, and quote later.
Voice note does not record or upload audio. Use the device keyboard microphone or paste a transcript. The result may suggest a summary, contact fields, follow-up, services, objections, and notes, but the representative must confirm information through the normal disposition, CRM, or scheduling workflow.
The current canvassing AI sheet does not include a photo-analysis control. Property photos and general estimate AI features remain separate from this canvassing-specific workflow.
Manager Mode
Owners and managers with both canvassing and AI access can open Manager AI analysis under Campaigns & territories.
Choose a campaign, optionally choose a territory, and ask a focused question such as:
- Which territory needs coaching attention?
- Where are we getting doors but not conversations?
- Which follow-ups are due?
- Which campaigns produced completed revenue?
- Which visits need GPS review?
The response is grounded in the scoped aggregate report. It shows Verified facts separately from Recommendations and marks recommendations for review before action. It does not infer a representative's motives or character.
Safety Boundaries
Every canvassing AI response follows these boundaries:
- Facts can cite only internal record identifiers supplied by the server.
- Customer notes are treated as untrusted text and sanitized before prompting.
- Logged input and output summaries redact email addresses, phone numbers, and secret-like values.
- Recommendations always require confirmation.
- The assistant cannot create or edit a campaign, visit, contact, lead, appointment, estimate, assignment, or setting.
- It cannot send a proposal, SMS, or email.
- It cannot invent measurements, prices, discounts, taxes, customer history, or guarantees.
- It cannot override deterministic estimate pricing or Red Line controls.
- Missing information should be stated instead of guessed.
The app shows Record-backed facts, Recommendations to confirm, and a reminder that no records changed. Complete any action in its normal workflow.
Privacy and Coaching Access
Field requests are restricted to the signed-in user's allowed territory or manager scope. Manager aggregate analysis does not automatically expose every private conversation.
By default, conversation-level coaching remains private while managers can review aggregate trends. Canvassing settings > Field policy > Manager access to private AI coaching stores the organization's policy preference, but the current manager AI screen remains aggregate and does not expose a conversation-level coaching archive.
Before enabling conversation-level manager access, notify representatives and confirm that the policy meets local privacy and employment requirements.
Connectivity and Usage
AI requires a network connection. When offline, the app disables AI controls and explains why; door dispositions, GPS capture, local syncing, and estimate drafts continue without AI.
Requests are rate-limited and use a client request UUID plus a fingerprint of the complete request. Before contacting the provider, the database atomically reserves a conservative credit amount and counts both reserved and completed usage against organization limits.
- An exact completed replay returns the saved response without another provider call or charge.
- A concurrent copy receives an in-progress response and retry window.
- After the provider succeeds, the result is durably staged before usage, billing, audit, and replay completion are finalized in one database transaction.
- If finalization is interrupted, a service-only reconciler finishes the staged result. The same request does not call the provider again.
- An uncertain provider transport outcome becomes indeterminate and is not automatically retried, because the provider may already have generated a result.
- Sensitive replay bodies expire after 24 hours and are removed by a bounded cleanup job.
If provider configuration, context lookup, quota reservation, or a definitive generation step fails, the request fails without changing canvassing or CRM records. A reconciliation or indeterminate message means the user should follow the displayed retry guidance or start a new request; it does not mean the assistant changed a business record.
AI output is assistance, not a customer commitment, property inspection, engineering opinion, insurance determination, or guarantee.
Was this article helpful?
Still need help? Contact support