How to Automate Self Storage Calls: A 30-Day Rollout Plan

# How to Automate Self Storage Calls: A 30-Day Rollout Plan

Here's the rollout that goes wrong, and it goes wrong the same way almost every time.

An operator signs with an AI vendor on a Thursday. Everything gets switched on the following Monday: payments, rentals, gate codes, move-outs, the works. By Wednesday a tenant has gotten a confusing answer about a rate increase and called the district manager to complain. By Friday the area manager has asked to "turn the AI off for now." Six weeks of evaluation, dead in nine days.

The technology wasn't the problem. The sequence was. Nobody decided which calls the AI should own first, nobody told the site staff what was changing, and nobody agreed in advance on what a good week would look like.

So here's a rollout plan that doesn't do that. Four weeks, one thing at a time, with a specific decision to make at the end of each.

Week 1: turn on the boring stuff only

Start with the calls where the right answer never changes and the stakes are low. Hours and directions. Unit sizes. What you need to bring to move in. Whether there's a gate code for after-hours access. Whether you take a lock or sell one.

These are unglamorous and they're the perfect week-one scope, for three reasons. The answers are objectively checkable, so you'll know immediately if something's wrong. Nobody's money is involved. And the volume is high enough that you'll have a real sample by Friday.

The practical work here is mostly writing down what your facility actually does. Most AI platforms give you a knowledge base for exactly this, and most operators discover during setup that two of their locations have been quoting different move-in requirements for a year. That discovery is worth the week on its own.

Decision at the end of week 1: read twenty transcripts. Not the summary metrics, the actual transcripts. Are the answers right? If two are wrong, fix the underlying knowledge base entry, not the AI.

Week 2: add payments, because that's where the volume is

About 1 in 3 calls to a self storage facility is about money. Someone's paying a bill, asking why the balance changed, or sorting out a card that didn't go through. This is the single biggest category on most books, and it's the one where automation earns its keep fastest.

It's also the one where "handled" has to mean something specific. A payment call is done when the card is charged and the payment is posted to your management system, not when the caller was sent a link and told to try it themselves. Before you turn this on, confirm with your vendor which of those two things is happening.

Across our own book, nearly 3 in 4 payment calls complete end to end. That number is a function of the integration underneath more than the voice on top, which is why it's worth proving in week two rather than assuming in week one.

Decision at the end of week 2: pull the payment calls specifically and check the posting. Did every charged card produce a payment record in your FMS? A gap here is the single most expensive thing to discover in month three instead of week two.

Week 3: decide your escalation policy on purpose

This is the week everyone skips, and it's the one that determines whether your staff trusts the system.

Some calls should reach a person immediately, and you should decide which ones rather than letting a default decide for you. The usual list: anyone who's angry about a rate change, anything involving an insurance claim or a break-in, any emergency, and anyone who asks for a human twice.

That last one matters more than operators expect. Some callers just want a person, and they don't owe you a reason. A system that argues with them produces worse outcomes than one that transfers cleanly on the first ask, and it produces much worse outcomes than one that transfers on the second.

Write the policy down. Then check that the handoff actually carries context: if a tenant spent ninety seconds explaining a problem to the AI and then has to start over with your manager, you've added a step instead of removing one.

Decision at the end of week 3: listen to five transferred calls. Did the human pick up knowing what the call was about?

Week 4: read the numbers by category, never blended

By now you have four weeks of calls, and you'll be tempted to look at one number: what percentage did the AI handle.

Don't lead with that. A single blended percentage is mostly a description of your call mix, not your system. Break it out instead:

Decision at the end of week 4: pick the one category performing worst and fix that specifically. Not the blended number.

> Rollout done wrong: Everything on at once. A bad answer in week one. Staff conclude it's unreliable. The blended number looks mediocre, nobody can say why, and by month two the conversation is about switching it off. > > Rollout done right: Week one proves the answers are correct. Week two proves the money moves. Week three means nobody at the desk feels ambushed. By week four you're not asking whether the AI works, you're asking which call type to improve next.

Same product, same facilities. Different order of operations.

The part everyone forgets: tell your staff first

The fastest way to lose a rollout is to let a site manager find out the AI exists because a tenant mentioned it.

Say three things, before week one. First, what it's taking over, and be specific: it's answering the hours and gate code calls so you're not interrupted mid-tour. Second, what it will never do: it won't handle an angry tenant or an emergency, those come straight to you. Third, how to report a bad call, with an actual name or channel, because your site staff will find problems before your dashboard does.

Skip this and the first weird call becomes evidence that management is replacing people. Do it and your managers become the best source of fixes you have.

Three things that actually break rollouts

Stale data. If the AI is reading from an overnight export, a tenant who paid at 9 AM is still delinquent at 2 PM. Every downstream answer inherits that error. Ask whether it reads live account status.

No owner. A rollout needs one person who reads transcripts weekly for the first month. Not a committee, not a quarterly review. Without it, small errors compound silently.

Turning on too much, too fast. Which is where we started. The work is not technically hard. It's just easy to sequence badly, and the cost of sequencing badly is the political capital you needed to finish.

What good looks like at day 30

You should be able to say, concretely: these four call types are handled end to end, these two go straight to a person by policy, this is the number for each, and this is the one we're working on next.

That's a different conversation from "the AI handles about half our calls." It's the difference between a system you're operating and a system you're hoping about.

Keep reading

Planning a rollout and want the four-week version mapped to your own call mix? [Book a demo](https://www.lumiostorage.com/book-demo) and we'll walk through what week one should cover at your facilities.