Almost every automation I have written about on this site assumes one thing quietly: that when the business sends a number, the customer is deciding about that number. A homeowner reads the roofing estimate. A patient reads the treatment plan. A bride reads the venue proposal. Moving companies do not get that. When a mover sends an estimate, the customer is reading it next to two other estimates, from two other companies, all quoting the same day, and none of the three numbers are constructed the same way.
That is the whole problem in one sentence. Movers rarely lose a job on move day. They lose it in the space between the quote request and the booked deposit, and they usually never find out why. As Gideon Wafula, AI Automation Engineer, I want to take that gap apart properly, because it is a different shape from the estimate-follow-up problem in the trades and it needs a different build.
There are four structural things going on here that do not apply to a plumber quoting a water heater.
The customer is running a bake-off by default. Consumer advice on choosing a mover has said the same thing for decades: get at least three written estimates, in person or by video survey, before you commit. That is not a bad customer being difficult — it is the standard advice, and it means your estimate is competing before it is even read. Any follow-up written as though yours is the only number on the table sounds tone-deaf.
The inventory is perishable and dated. A crew-day on the last Saturday of the month either sells or evaporates. You cannot store it, discount it later, or move it to a slower week. That is the same economics I wrote about in the seasonal pre-booking piece, except compressed: the customer has a lease end date, a closing date, or a landlord, so their timeline is fixed and short.
The estimate itself is a scheduled appointment that can no-show. Most reputable movers now do a video walkthrough or an in-home survey before quoting. That survey is a booked event with a start time, and a meaningful share of them are simply not attended. So there is a no-show problem sitting inside the sales process, before any quote exists.
The conversion event is a deposit, not a signature. This is the one people get wrong. A signed estimate holds nothing. The job is booked when money moves. And deposits in moving are constrained — the widely repeated consumer guidance is that a mover should not be asking for much more than a small percentage of the estimated cost up front, and outright large prepayments are a documented red flag for scam operations. So your automation is chasing a small, awkward, trust-dependent payment from someone who has been told to be suspicious of movers asking for money. That is a very specific writing problem.
Before building anything, export the last twelve months from your CRM and answer five questions. Not blended averages — segmented, because the average is always a lie here.
Then call twenty people who got a quote and booked somewhere else. Ask them one question: what made you pick the other company? You will hear four answers — they were cheaper, they answered first, I could not tell what your price included, and I never heard from you again. Only two of those are automation problems. The published case studies of movers who put structured follow-up in place tend to report booking rates moving from roughly the high twenties into the high thirties or low forties on the same lead volume, which is a large swing, but it is worth noting that those numbers come from CRM vendors describing their own customers. Treat them as a direction of travel, not a promise.
I use n8n for this, sitting on top of whatever moving CRM the company already runs — SmartMoving, Supermove, Movegistics, MoveitPro, or in smaller shops a spreadsheet and a shared inbox. Six steps.
A lead lands from any source — website form, marketplace, Google, phone. Within seconds, an SMS goes out that does three things: confirms the request was received, names the specific move (origin city to destination city, and the requested date if it was captured), and states when a human will call. Nothing else. No price, no pitch, no "we're the best movers in the tri-state area."
The reason this works is not persuasion, it is placement. The customer submitted forms to three companies in the same ten minutes. The first company to appear on their phone is now the one the other two are being compared against. I covered the underlying mechanics in detail in the five-minute window post; the moving-specific twist is that naming the origin and destination in the first message is what stops it reading as an autoresponder.
The acknowledgement carries a live booking link for the video survey. Once a survey is on the calendar, a separate short sequence protects it: a confirmation at booking, a reminder the day before, and a reminder an hour ahead with the join link and a one-tap reschedule option. Give people an easy reschedule and a chunk of would-be no-shows become moved appointments instead of dead leads.
This step is deliberately dumb. No model, no personalisation beyond name, date and link. It is a reliability problem, not a copywriting problem, and every clever thing you add here is a new way for it to break.
When the estimator finishes, the quote goes out with the inventory, hours, crew size, truck count, and what is and is not included stated plainly — and critically, with the type of estimate named in the customer's own language. Non-binding, binding, or binding not-to-exceed are not interchangeable words, and in the US they are regulated terms with specific meanings for how much a customer can be charged on delivery day. A customer comparing three numbers with no idea that one of them is a ceiling and another is a guess will just pick the small one.
Explaining the structure of your price is the single most useful thing you can put in an estimate follow-up, and it is the thing almost nobody does. The general pattern here is the same as the unsold estimate follow-up build — the difference is that in moving, the follow-up's job is comparison literacy rather than urgency.
Three touches over roughly a week, timed to your own measured quote-to-deposit window rather than to a generic drip schedule. Touch one, a day after the quote, offers to walk through what the number covers. Touch two, a few days later, is the only one that mentions the calendar, and it says something true and specific: the requested date is held provisionally and other enquiries are asking for it. Touch three closes the loop and gives explicit permission to say no, which sounds counterproductive and is the message that most often gets a real reply.
Stop the entire sequence on any inbound message. Always. A customer who replies is now a human's job.
When the customer says yes, the deposit request goes out immediately with the amount, what it secures, the cancellation terms, and a payment link — as a plain confirmation of something already agreed, not as a sales push. Two reminders maximum. If it is unpaid after that, a human calls, because an unpaid deposit after two reminders is almost never forgetfulness. It is usually a second thought, a partner who has not agreed, or a competing quote that arrived late.
Two guardrails matter here. Never let the model generate the amount, the percentage, or the cancellation terms — those come from stored fields. And never escalate the tone. A mover pressing hard for money before the job triggers exactly the alarm consumers have been trained to listen for.
This is the step most movers skip and it is where the money is. When a deposit is not paid and the quote goes cold, the held date has to actually be released back into inventory and offered to the next-best quote sitting on the same date. That means a nightly job that identifies dead holds, frees the crew-day, and notifies sales which live quotes match it. Without this, your automation politely follows up on lost deals while a sellable Saturday quietly expires.
The language model writes tone, handles the messy free-text of an inbound reply, and classifies intent. It does not produce numbers. Every price, deposit amount, date, crew size, inventory line, mileage figure and cancellation term is injected as a validated variable from the CRM record, and any draft containing a currency figure the system cannot trace back to a stored field gets rejected before it sends and falls back to an approved template.
This is stricter than it sounds and it is not optional. An estimate is a document with legal weight, and in interstate moves it sits inside a federal regulatory framework. A model inventing "we can probably do it for around $2,400" in a follow-up text has not written bad marketing copy; it has made a representation about price that a customer will screenshot. If you take one thing from this post, take that.
The other guardrails: send-time suppression rather than selection-time, covering already-booked customers, cancelled moves, open complaints, human contact in the last twenty-four hours, opt-outs and duplicate household records; TCPA consent, opt-out honouring and local daytime hours for US SMS; and for UK and EU work, the usual lawful-basis separation between servicing an enquiry someone actually made and marketing something they did not ask for.
Running cost lands around 30 to 120 USD per month for messaging, model calls and the automation platform. The real spend is one-time: CRM integration, fixing lead capture so origin, destination and requested date are reliably present, and building the human escalation path. For a company booking jobs in the low thousands, one additional booked move a month pays for a year of operation, which is why this is an easy business case to make and a hard one to fake — the metric is visible in the CRM.
Baseline four numbers and watch them: median time to first human contact split by staffed hours; survey held rate; quote-to-deposit rate by lead source against its own pre-build baseline; and opt-out and complaint rate as an automatic circuit breaker that pauses the workflow if it crosses a threshold.
And the honest caveat, same as every build I write up. If you are losing quotes because you are genuinely more expensive and cannot explain why, faster follow-up gets you rejected sooner. That is still useful — a fast no frees the crew-day while it can still be sold — but it is not the outcome anyone is hoping for. Fix the pricing story first, then automate the gap.
The same structure ports cleanly to junk removal and hauling, piano and specialty transport, auto transport brokers, commercial office relocation, and any business where a dated crew slot is sold against two competing quotes. If you want the full range of what I build around this pattern, that is on my AI automation services page.
Gideon Wafula builds custom AI automation systems, n8n, WhatsApp, Voice AI, and more.
See Services →