Few operators can say what share of their calls actually get answered. The ones who measure it are usually surprised, and not in a good way.
Cedar Creek Capital measures it. The company runs 22 self-storage facilities, about a third of them fully remote, with a call center staffed by four agents. Across three different approaches to answering the phone, one number stayed constant in the reporting: the percentage of inbound calls that got picked up.
Here is what that number did.
Humans alone: 60 to 65 percent. Four agents, phones that did not stop, a growing portfolio.
Another AI solution: low to mid 80s. A real improvement that then stopped improving.
Alita AI Chat and Voice: about 94 percent. Across 21 facilities.
Three setups, three plateaus. What separates the last two is worth understanding, because that is the gap most operators are about to walk into.
Sixty to sixty-five percent is not a failure of effort. It is what four people can do against a call volume that never pauses, in a portfolio that keeps growing.
The calls that go unanswered are not evenly distributed either. They cluster at the times when everyone calls at once, which is also when the callers are most likely to be prospects deciding between you and the facility down the road.
Cedar Creek set a 90 percent company goal and chased it for more than a year without reaching it. With significant expansion ahead, close enough was not going to scale.
Cedar Creek brought in an AI voice solution before Alita, and it worked. The answer rate moved from the low 60s into the low-to-mid 80s, a real gain on a number that had been stuck for years. Then it stopped, about six points short of the goal, and stayed there.
Nearly every AI vendor in storage says its agent reserves spaces, takes payments, and issues gate codes. Most of them can. That claim no longer separates anyone, which is why a measured comparison between two of them is worth more than either one's marketing.
The limits were structural. It saw only a slice of the account, sitting next to the operating data rather than inside it, so any caller whose question needed the full account got handed to a person anyway. Its configuration ran through whatever fields the vendor exposed, which is not the same thing as knowing how a particular facility runs. The last ten points lived in the parts of the account it could not reach.
Most AI agents in this category connect to a facility management system through an integration. Alita is part of Hummingbird, which sounds like an architecture footnote until you look at what it changes day to day. Alita reads the same live data the team already manages, so when someone changes pricing at a facility, Alita has the new pricing, and there is no sync to break or second system for anyone to remember. What it knows about each property came from Cedar Creek: their pricing, their rental procedures, their billing policies, their move-out processes, facility by facility, rather than a set of fields a vendor decided to expose.
And it finishes calls. Quotes a real price on a real space, reserves it, sends the payment link, texts the gate code. A tenant who calls at nine on a Saturday needing a gate code gets the gate code.
The result was about 94 percent of calls answered across 21 facilities, and a 90 percent goal that the company had been chasing for over a year.
As Jesse Harmon, VP of Marketing, Revenue, and Commercial Strategy at Cedar Creek Capital, put it: "We're at almost a 94% answer rate. I'll take that all day long." most operations, each edge already has a workaround, and the workarounds are worth taking seriously.
Market data comes from a subscription, a revenue management tool, or a manager who knows the market well. The new question gets answered in a spreadsheet somebody assembles from two or three exports. The analyst builds their view by exporting on a schedule and rebuilding it each month.
What a workaround cannot do is scale, survive turnover, or run at the moment a decision gets made. It runs when somebody has time. And because each answer costs effort, a question only gets asked when it is urgent enough to justify the work. Plenty of useful ones never clear that bar.
That is the real cost at the edges. Not a lack of information, but information that arrives on a schedule set by how busy somebody is.
This is the question every owner asks second, right after the answer rate.
Cedar Creek's call center went from five agents to three. Both reductions happened through natural attrition, with zero layoffs and no backfill planned. The agent who used to handle nothing but leads now helps run management and revenue operations.
Facility count kept growing over the same period, absorbed by the team already in place.
The same people are doing work they could not get to before, because the routine calls resolve themselves.
If you take one thing from Cedar Creek's experience, make it this: measure your answer rate before you evaluate anything.
Most operators cannot say whether they are at 60 percent or 85 percent, which makes every vendor claim impossible to judge. Once you know your own number, two questions sort the market quickly.
What share of answered calls end in a completed action? Not a transcript, not a lead, an actual reservation, payment, or gate code issued. Every vendor will say yes to this one, so ask for the number itself, by facility, from a deployment you can call and verify.
What does the system see? An AI reading live operating data can resolve an account question. An AI reading a copy, or a subset, will hand it back to your team. That difference is where Cedar Creek's last ten points came from.
Operators at SSA Fall asked a longer version of this list, and the answers are worth having before you sit through a demo: What Operators Actually Asked About AI at SSA Fall 2026.
Alita AI Chat is live on Mariposa websites now. Alita AI Chat and Voice, which adds the phone line, is available and rolling out to operators now.
Everything above ran on the first two.