AI Photo Moving Surveys Sold From Abroad

Disclosure: this article contains affiliate links. If you open an account through one of them, Cashflow Abroad may earn a referral commission at no extra cost to you.

One 2,000-pound miss on a non-binding interstate estimate can cap what a mover collects at delivery. Under 49 CFR 375.401, a non-binding figure that is too low is not a rounding error; the shipper can take the goods after paying 110 percent of that estimate. Independent operators still try to skip the survey because an in-home visit costs a dispatcher an afternoon and a tank of diesel. The offer you sell from abroad is the replacement: a photo-and-video inventory, mapped to cube sheets, QA'd in under four business hours, so the owner can send a written estimate the same day the lead lands.

This is not a generic chatbot gig. You sell a survey desk to 2-to-8-truck US independents who still quote from memory and lose the race to three other companies the customer called the same morning. It sits in the AI Income & Cash Flow cluster: dollar invoices, a thin tool bill, and delivery you can run from any time zone with a Mercury Bank operating account.

What you sell, and who actually pays for it

Package the work as a named product, not "AI help." The buyer is the owner-operator of a licensed household-goods carrier or a local mover who also runs the occasional interstate job through an affiliated certificate. They already pay for leads. They do not have a full-time estimator. Their close rate sits in the low 20s when quotes take a day; the shops that answer with a room-by-room inventory first take a disproportionate share of bookings.

What leaves your desk is not a dollar price. Pricing is the mover's tariff problem. You deliver: a room-by-room item list, industry cube and weight lookups, accessorial flags (flights of stairs, long carry, packing, piano, tight driveway), missing-photo requests, and a one-page estimator brief the owner can paste into their estimate form.

Three SKUs that close without custom scoping

SKU	What the mover gets	Typical price	Your delivery time
Launch build	Branded photo-upload link, SMS consent copy, Airtable base, cube table, 10 sample surveys	$1,200–$1,800 one-time	8–12 hours over 7 days
Survey retainer	Up to 30 completed inventories per month, 4-business-hour SLA, Slack/email brief	$450–$650 / month	~15 minutes QA each
Burst / storm week	Overage surveys, weekend coverage, extra missing-photo chases	$14–$18 per survey	Same SLA, billed weekly

Do not sell "AI that sets the customer's price." FMCSA rules put the written estimate on the carrier. Your product is the physical survey layer: seeing the goods, counting them, and handing the owner a cube sheet they will sign. Local-only jobs are still worth selling; many independents never touch interstate tariffs and still waste two days driving to three-bedroom houses that were never going to book.

Do photos count as the FMCSA physical survey?

Yes, if the mover can actually see the goods and a human reviews the result before it becomes an estimate. Appendix A to 49 CFR Part 375 tells shippers a physical survey is "conducted on-site or virtually," as long as the mover sees the household goods. The shipper may waive a survey in writing, but waivers are a claims magnet. Your sales pitch is that a reviewed photo walkthrough is the virtual survey, not a loophole around it.

That distinction is why this retainer survives a bad week. Yembo, MoveBlueprint, and HomeSurvey.ai already sell video-survey SaaS to the same buyers. Independents balk at another $99–$199/month tool they will not learn. You sit beside those products: either you operate their existing tool, or you run a thinner stack they own (upload link + your QA). Competing on "we have AI" is a dead end. Competing on "your estimator gets a cube sheet before dinner" is cash flow.

Industry cube tables, not the model "measuring" a dresser from a phone photo, should drive volume. Vision models misread particleboard as solid wood and miss closet contents. You map detected items to a published cube sheet the mover already uses, then flag uncertainty instead of inventing pounds. If the job will later convert cube to weight, section 375.401(e) requires the carrier to explain that formula in writing. Put a reminder on every brief so the owner does not skip it.

Tool stack and what it actually costs

As of September 2026, you can stand this up without a custom app team. n8n routes the webhook. Airtable is the job board. Anthropic or OpenAI does first-pass item ID. Twilio sends the upload link after the lead texts in. A $6 DigitalOcean Droplet hosts a one-page uploader if you do not want photos sitting in a form product. Keep API keys on accounts the mover owns when you can; it makes offboarding clean and matches how owner-owned agents are sold.

Monthly bill for a three-client desk

Tool	Plan to start	Published price (Sep 2026)	Role
n8n Cloud	Starter (2,500 executions/mo)	€20/mo billed annually (~$22)	Webhooks, SMS, Airtable writes
Airtable	Free, then Team 1 seat	$0, then $24/seat billed monthly ($20 annual)	Jobs, photos, cube output, SLA clock
Twilio SMS / number	US long code + A2P 10DLC	$0.0083/segment; number $1.15/mo; brand+campaign fees below	Upload-link texts, missing-photo chases
Anthropic Claude	Haiku 4.5 + Sonnet 5 QA	Haiku $1 / $5 per 1M in/out; Sonnet 5 $2 / $10	Item ID, then a second-pass exception review
OpenAI (backup)	GPT-4.1 mini	$0.40 / $1.60 per 1M in/out	Failover when Anthropic queues
DigitalOcean Droplet	1 GiB / 1 vCPU / 25 GiB	$6/mo	Plain photo uploader, no UI text on stored images

Data note: n8n Starter and Pro list prices on n8n.io/pricing are €20 and €50 per month when billed annually; monthly billing is higher for the same execution caps. Twilio SMS list rates are on Twilio's US SMS page. Model rates change; lock them in your client contract as pass-through.

Startup cash, not including LLC paper if you already have a US entity: A2P 10DLC standard brand registration is $44 one-time plus a $15 campaign vetting fee on Twilio's published 10DLC page; low-volume and sole-proprietor brands are $4 plus the same $15 vetting. Campaign monthly fees run $1.50–$10 ($2 for sole proprietor). Add $20 of API credit and a weekend of build time. Total cash outlay to first paid survey is typically $90–$150 plus your hours.

Route client payments into Mercury Bank so USD retainers do not sit in a personal checking account you will later struggle to explain on a visa file. Charles Schwab remains the ATM and brokerage side of the stack; do not mix operating deposits with brokerage cash.

Fulfillment workflow you can run from UTC+7

The mover's customer never needs to know you live abroad. They get a text from the mover's number, a branded upload page, and a call from the owner with a real inventory. You work the queue in Airtable views: New, Waiting on photos, Ready for QA, Sent. Overlap 6–10 hours with US afternoon so 4-hour SLAs do not slip into the next calendar day.

Build week: nine steps, then stop adding features

Step	Action	Done when
1	Collect tariff cube sheet or a standard household-goods cube table the owner already uses	CSV with item name, cu ft, lb
2	Stand up Airtable: Leads, Rooms, Items, Flags, Audit log	One base per client, client login read-only
3	Register A2P 10DLC on the mover's EIN when possible; use their Twilio subaccount	Campaign approved, number in a Messaging Service
4	Write consent language: informational survey texts, STOP, frequency, not a quote	Counsel-reviewed if they do outbound marketing; keep yours transactional
5	Deploy the uploader: room buttons, 15–40 photos, optional 60-second walkthrough	Test lead completes on a phone in under 8 minutes
6	n8n: inbound SMS keyword or form → create Lead → send link → ingest files	One execution per survey, not per photo
7	Vision pass: Haiku lists items per photo; script maps to cube CSV; Sonnet reviews mismatches	JSON with confidence <0.6 rows highlighted
8	Human QA: add missing closets, packing, wall-mounted TVs, safes, gym gear	Brief PDF/HTML emailed to owner within SLA
9	Missing-photo chase: one SMS, then stop. Owner calls if still incomplete	Logged; never three-text the customer

Developer prompt that belongs in the workflow, not in a blog comment: "You are a moving estimator's clerk. Identify only visible household goods. Map each item to this cube table. Never invent weight. If a closet, garage, or shed is not photographed, add a MissingPhoto flag. Never output a dollar price or crew count." Structured outputs beat chat. Store the raw model JSON in the audit log so a claim six months later has a file, not a vibe.

US-side demand testing does not require a website. Post the offer as a free listing on Brixaz in two metro areas the mover already serves, framed as "photo moving survey / virtual estimate help," and see whether owners reply before you spend a week on a cold-email sequence. Listings are free; treat it as a classifieds probe, not a growth engine.

If you already sell other AI retainers, this desk pairs cleanly with the same n8n muscle as an n8n automation service. Keep the moving SKU separate on invoices so a mover can cancel surveys without pulling down unrelated workflows.

How to price it without racing SaaS

SaaS video survey tools advertise $20 per survey or $99–$199/month unmetered. You will lose a bake-off against "unlimited AI" if the owner is software-curious. You win when the owner is software-averse and already drowning. Price against the cost of a part-time estimator, not against Yembo. A dispatcher burning two hours per in-home at $22/hour is $44 of labor plus fuel before a single booking. Your $550 retainer covering 30 surveys is cheaper than 12 wasted drives.

Close on a 14-day paid pilot: $400, 12 surveys, SLA relaxed to 8 hours, then convert to retainer. Require the owner on a 30-minute tariff walkthrough so you learn their accessorial list. Do not discount below $12/survey blended; that is where QA quality dies and you inherit underpriced jobs as "your AI was wrong."

Expats who treat this like Upwork gig work underprice and churn. A tighter packaging path is in the AI freelancing formula: named offer, dollar retainer, delivery hours on a calendar. US persons still file, and self-employment tax follows you abroad even when Form 2555 zeros income tax.

What goes wrong, and how you stay off the claim

The failure mode is not a missed sofa. It is a binding or non-binding estimate that undercounts because the model skipped a basement, the customer hid a treadmill, or you labeled particleboard as oak and the owner priced a luxury move. Write every brief with three buckets: Counted, Uncertain, Not photographed. Uncertain items never become confident cube. Not photographed items never become "empty rooms."

TCPA still requires prior express consent for autodialed marketing texts to mobiles. Informational "here is your upload link" messages after the customer asked the mover for a quote are a different fact pattern than blasting specials. Put STOP on every SMS. Register 10DLC. Do not text from your personal foreign SIM. The Eleventh Circuit vacated the FCC's 2023 one-to-one consent add-on in January 2025; that does not waive written consent for telemarketing. Keep your traffic inside the mover's existing lead conversation.

Other landmines: interstate jobs without the carrier's USDOT/MC on the estimate; you drafting language that looks like you are the broker; storing kids' rooms and document piles in a personal Google Drive; promising "FMCSA compliant" as if you were their lawyer; weekend SLAs you cannot staff during peak May–August season. Cap each client at a survey count you can QA at 15 minutes. When the queue hits 20 unreviewed jobs, stop intake. Underpriced accuracy is how movers get stuck at 110 percent of a bad number.

Data handling: photos are the customer's home. Use the mover's Twilio and object storage when they will pay the $6 Droplet. Sign a one-page processing addendum. Delete raw images on a 60-day timer after delivery unless the mover's claim policy needs longer. Do not train public models on client photos.

Conclusion

Independent movers do not need another demo of a talking avatar. They need a cube sheet before the competitor's estimator gets in the car. Sell the survey desk, not the tariff. Keep the stack under $80–$100 a month until the third client, QA every inventory, and let FMCSA's virtual-survey language do the sales work that "AI" cannot. Geographic arbitrage is the margin: US retainers, a low-cost time zone, and a bank account that still looks like a US operator to the customer writing the check.

Data notes / Sources checked

• 49 CFR 375.401 (eCFR, current through Sep 2, 2026) — physical survey, binding vs non-binding, volume-to-weight explanation
• 49 CFR Part 375 Appendix A — "on-site or virtually" survey definition for shippers
• Twilio US SMS pricing and A2P 10DLC fees
• n8n Cloud pricing
• Anthropic Claude API pricing
• OpenAI API pricing
• DigitalOcean Droplet list prices (Basic 1 GiB at $6/mo as of 2026) checked on the vendor pricing page; do not paste referral URLs in-article
• Airtable pricing
• IRS self-employment tax (15.3%) and FEIE / self-employment interaction