BundleBox Survey
AI surveys that turn [X] minutes of video into a priced-ready inventory, whichever of four ways the home is surveyed.
Role
Product Designer
Skills
AI Product Design · Interaction Design · Systems Design
Timeline
2 Weeks
Team
2 Product Designers
BundleBox Survey is an AI-powered virtual survey product for moving companies. Instead of sending an estimator to every home, movers can survey a move four ways: a self-recorded video, a live assisted video call, an on-site visit, or an uploaded video. In every case, AI turns the footage into a room-by-room inventory with volume and weight.
I designed Survey v2 end to end. The work covered survey creation, AI transcript and inventory review, the inventory workspace, customer-facing inventory review, automated customer communication, and the dashboards teams use to run surveys at scale.
Overview
Product walkthrough: from the dashboard to a generated inventory
Problem
Accurate moving quotes depend on knowing what’s in the home. Traditionally, that meant an estimator driving to every address, walking each room and counting items by hand. Early virtual surveys removed the drive, but they produced unstructured video that someone still had to watch and transcribe into an inventory.
Current process of surveying a home
Every survey type followed a different process, and customer follow-up was manual. Surveys stalled when customers didn’t record, links expired unnoticed, and the inventory that finally came out lacked the details crews and estimators actually needed, such as packing, fragility, disassembly and crew instructions.
A single in-home survey cost an estimator [X hours] between travel, the walkthrough and retyping the inventory.
Video surveys captured the home but not the data. The footage still had to be manually converted into items, volume and weight before anyone could price the move.
Target Users
The primary users are surveyors and estimators at moving companies. They schedule and run surveys, review AI-generated inventory, and hand accurate data to sales and dispatch. Office admins oversee survey volume and follow-ups across the whole team. Customers are secondary users: they record their homes, join video calls, and review their inventory before signing.
Meet Rick, our primary user
Surveyor at a [size]-truck mover, running [X] surveys a week.
Survey more homes without driving to them, and trust the inventory enough to price from it.
Customers who never record, scrubbing through long videos, and inventory that misses packing and fragile items.
Phone, video calls, spreadsheets, manual texts and emails.
Problem Statement
Discovery Insights
To understand why surveys stalled and why inventories weren’t trusted, we mapped every survey type from creation to confirmed inventory. We did this through interviews with surveyors and admins, and a review of real survey footage.
Survey lifecycle from invite to confirmed inventory, with drop-off points
We spoke with [X] surveyors, estimators and admins across [Y] moving companies and reviewed [X] recorded surveys. The goal was to find where customers dropped off and what an inventory needed to be usable downstream.
Key Insights
Some customers want to record on their own time, others need a live call, and some homes still need an in-person visit.
Customers rarely refused a survey. They forgot, ran out of time, or let the link expire.
Crews and estimators needed volume, weight, who packs each item, fragile, high-value and bulky flags, disassembly needs, and crew notes.
Surveyors only trusted a detected item when they could jump to the moment it appeared on video.
These insights shaped a core design principle:
Strategy & Key Decisions
Based on research into how surveys were run and where they broke, we made two structural decisions that shaped Survey v2.
Survey creation became a single flow where the surveyor picks a capture mode (self virtual, assisted virtual, on-site or upload video). Customer details, move details and the outgoing SMS or email adapt to that mode. However the survey is captured, it ends in the same transcript and inventory.
Why: Customers get the format that works for them, while surveyors and downstream teams always receive the same structured output.
The AI inventory was designed as a working document. Items are grouped by room, with quantity, volume, weight, packing responsibility, handling flags and crew notes. A live summary shows totals per room, and the inventory can be exported as CSV or PDF or sent to the customer to review.
Why: An inventory only saves time if estimators can price from it and crews can work from it without re-entering anything.
Risks & Trade Offs
Detected items could be wrong or assumed. Every item links to a video timestamp, inferred items are labelled (for example, “Not mentioned by customer, considered moving by default”), and surveyors can switch models and retry processing.
Automated reminders could quickly become spam. Each message is tied to a specific trigger with defined timing, such as 24 hours after invite or 24 hours before link expiry. Templates are editable and can be sent as a test first.
Admins needed an organisation-wide view while surveyors needed their own queue. The dashboard switches between organisation and personal scope, and every inventory edit is logged in Activity with who made it.
Solutions
One create flow covers all four survey types. Scheduling, customer details and the outgoing SMS or email adapt to the mode, and the message can be edited in place.
Self virtual survey, filled in
Assisted virtual survey
On-site survey
Calendar overlay
Switching the message to email
Before: A separate setup for each survey type, with messages written by hand. [Add old design screen from Survey v1 or the MVP.]
AI transcribes each video by room and detects items with timestamps. Surveyors can review what was said, see what was detected, and jump to any moment in the footage.
Survey tab with several surveys
Assisted virtual survey
Processing an uploaded video
Transcript and AI inventory
Send reminder modal
Before: Scrubbing through raw video to note items by hand. [Add old design screen from Survey v1 or the MVP.]
A room-grouped inventory with volume, weight, packing responsibility, handling flags and crew notes. Totals update live and the inventory exports as CSV or PDF.
Inventory tab
Summary expanded
Crew instructions and internal notes
Video modal
Zero state
Filter dropdown
Before: A flat item list with no volume, weight or handling details. [Add old design screen from Survey v1 or the MVP.]
Customers review their room-by-room inventory on any device before signing the Bill of Lading, which catches disputes before moving day.
Customer inventory on web
Customer inventory on mobile
Flag filters close-up
Before: The customer sees their inventory for the first time on moving day. [Add old design screen from Survey v1 or the MVP.]
Automated SMS and email for every survey stage, including invitations, reminders, expiring links, rescheduling and submission. Each message uses editable templates, variables and a branded sender identity.
Triggers overview
Templates
Variables
Sender identity
Send a communication
Before: Reminders sent manually, or not at all. [Add old design screen from Survey v1 or the MVP.]
These screens cover the rest of the product: dashboards, survey and customer management, and an activity trail that keeps teams in sync.
Admin dashboard
User dashboard
Needs Attention table
Survey list
Survey details with Zillow property data
Activity log
Customer list and card views
Impact
[X]%
Fewer in-person surveys
Or [X] hours of estimator drive time saved per week.
[X] min
From video upload to generated inventory
Time from a customer’s upload to a room-by-room inventory with volume and weight.
[X]%
Survey completion rate
Completion after automated invitations, reminders and expiring-link nudges.
4
Survey modes, one inventory
Pre-launch fallback: In development · [X] automated communication triggers · [X] pilot movers.
Learnings
Surveyors adopted AI-detected inventory only when every item could be traced back to the video and edited with a visible history.
Supporting four capture modes only worked because they all converged on one inventory model. The flexibility is upstream, and the structure is downstream.