AI After-Hours Support for NZ Self-Storage

It is twenty to ten on a Sunday night in Porirua. A tenant has backed a ute and trailer up to the gate of a storage facility, with a couch to drop off before a Monday move. He punches in his code. The keypad blinks red. He tries again, slower. Red.

So he rings the number on the sign. What happens next depends on who, or what, picks up, and on one question the facility has to answer long before that call comes in: who is allowed to decide that this person gets through the gate tonight?

Self-storage is a quiet business during the day and a surprisingly busy one after dark. Most calls outside office hours fall into four groups, and each one wants a different response. We build voice agents for this kind of work. Below we walk through what those four calls look like, where the line sits between what an agent handles and what a person decides, how to treat genuine emergencies, and what you need in place before any of it is worth switching on.

Close (ESC)

What do after-hours calls to a storage facility look like?

Four kinds, in most facilities: tenants locked out at the gate, tenants whose access has been suspended for an overdue account, people wanting to rent a unit, and genuine emergencies such as an alarm, a break-in or water coming through a roof. The mix varies by site, so the first job is to count your own.

We are not going to quote a percentage split here, because we have not seen one we trust. A facility beside a motorway with 24-hour access gets a different phone from a rural site that closes its gate at 7pm. A month of your own call log, sorted into those four buckets, will tell you more than any industry average.

What the four have in common is timing. They arrive when the office is shut and whoever holds the after-hours phone is at dinner, asleep, or at their kid's netball.

Why do tenants get locked out at night?

Three reasons come up again and again. The tenant has forgotten or mistyped a code they use twice a year. The keypad has locked them out after too many wrong attempts. Or their account is overdue and the access system has suspended their code automatically.

That last one catches people off guard. Management platforms such as Storman integrate with access control systems from vendors like PTI and ICT, and Storman describes the feature plainly: overdue customers are locked out automatically and let back in when they are up to date. It is a sensible rule. It also means the tenant at the gate often has no idea their card expired last month.

There is a fourth, less dramatic one. The tenant has arrived after the site's access hours have closed, and the code is working exactly as designed.

What does a human answering service cost for this?

Not a great deal, which is worth knowing before anyone pitches you on savings. Published NZ pricing for message-taking services starts from around $19 to $49 a month, with per-call or per-minute charges on top. A facility taking a handful of after-hours calls a week would likely spend in the low hundreds each month. Those figures are indicative, so check them against your own provider.

The limitation is reach. An answering service can take a clear message and pass on a phone number. It cannot see the tenant's account, cannot tell whether the lock is a forgotten code or an overdue balance, and cannot send a payment link. The tenant gets a courteous callback promise and stays on the wrong side of the gate.

Where should the line sit between the agent and a person?

The agent handles information and anything reversible. A person decides anything that changes who can physically get onto the site. Issuing a code, lifting a suspension manually or opening the gate remotely are security decisions, and they belong to someone accountable for the facility.

Close (ESC)

That boundary is the whole design. Get it right and the agent takes most of the after-hours load with very little risk. Get it wrong and you have built a polite, tireless way for a stranger to talk their way into someone else's unit.

On the agent's side sits everything that is pure information transfer: explaining access hours, telling a tenant how to reset a code through the facility's app or portal, taking a detailed message, answering questions about sizes and prices, booking a viewing, and deciding which calls are urgent enough to wake someone. On the person's side sits anything involving the gate, the lock, a dispute, or discretion.

Can an AI agent give out a gate code?

Our default answer is no, and it is the rule we would build to. An agent should never read a gate code aloud to a caller. A name and a unit number are easy to find out, and caller ID can be spoofed, so what someone says on the phone is weak proof of who they are.

What an agent can do, where the access system supports it, is trigger a code reset that goes to the mobile number already on the tenant's file. Proof then comes from possession of that phone, and the agent never handles the code at all. If the access system cannot do that, the agent explains the options, logs the call, and passes anything urgent to the on-call person.

This is one place where being unhelpful in the moment is the right call. A tenant who waits until morning is annoyed. A tenant whose unit was emptied because an automated system trusted a confident voice is a much bigger problem, for them and for you.

What about a tenant locked out for an overdue account?

Once the caller is confirmed as the tenant, the agent can explain that access is on hold because of an outstanding balance and send a payment link to the mobile number on file. Where the management software restores access automatically once payment clears, the tenant's normal code starts working again without the agent touching the gate.

The agent does not negotiate. Payment arrangements, disputed charges, and anything to do with the lien and sale process under your storage agreement stay with a person during business hours. Those conversations carry legal weight and a real relationship, and nobody wants them handled by a script at midnight.

How should the agent confirm who it is talking to?

Treat caller ID as a hint and nothing more. Before discussing an account, the agent sends a one-time link or code to the mobile number on file and asks the caller to confirm it. If the number does not match an active tenant, the agent does not confirm whether a unit, a name or an account exists.

This matters for privacy as much as security. Principle 5 of the Privacy Act 2020 expects businesses to take reasonable safeguards against personal information being disclosed to the wrong person. Telling a stranger that "unit 42 is three months behind" is a disclosure. So is confirming that a particular person rents there at all.

How should emergencies be handled after hours?

Life safety comes first. If anyone is in danger or a crime is in progress, the agent tells the caller to ring 111 and keeps the rest of the call short. After that, its job is to reach a human fast: the on-call manager first, then a backup, with a written summary sent to both.

Close (ESC)

An agent does not replace a monitored alarm. If your site already has back-to-base monitoring, that stays exactly as it is. The agent covers the calls that arrive by phone: a tenant who can hear the alarm going, a neighbour who has seen the fence cut, someone who has noticed water pooling under a row of doors.

We wrote a longer piece on designing the moment an AI hands a call to a person, and the same rules apply here. Tell the caller what is happening. Give the human the context so nobody has to repeat themselves. Never leave a caller in silence while the system works something out.

Which calls should wake someone up?

A short, written list, agreed in advance. A break-in or suspected break-in. An alarm sounding. Flooding, fire or smoke. A tenant still inside when the gate has closed for the night. A gate stuck open, since that is a security problem for every unit on site.

Equally important is the list of calls that wait until morning: forgotten codes outside access hours, move-in enquiries, billing questions and general complaints. The agent handles those, logs them, and leaves them for the office. An on-call phone that rings for everything stops being answered, and then it misses the one call that mattered.

What happens if nobody answers?

The agent works down a fixed chain: on-call manager, backup, then owner, calling each and sending a text summary. It tells the caller it is trying to reach someone and gives an honest time frame. Every attempt is logged with times, so the morning review shows exactly who was contacted and when.

Keep the chain short and the roster current. On-call lists drift quietly, and a chain that ends at the number of someone who left six months ago is no chain at all.

Can an agent handle move-in enquiries at night?

Yes, and it is one of the better fits. People shop for storage in the evening, often in the middle of a move or a break-up or a renovation. An agent can explain unit sizes, quote from your current rate card, confirm access hours, answer what can and cannot be stored, and book a viewing or send a link to reserve online.

The numbers make this worth doing. Indicative NZ rents for 2026 run from around $80 a month for a small locker to roughly $160 to $195 for a 3x3 unit in a regional centre, with Auckland higher. A single unit let at $170 a month is about $2,000 a year, and many tenants stay well beyond the first month.

Keep the agent to your published prices. Discounts, move-in specials and anything off the rate card stay with a person, who can decide case by case.

What does the agent need from your management software?

Less than you might think. Read access to unit sizes, rates and availability. A way to match a caller's number to an active tenant and see a simple account status: active, overdue or suspended. And somewhere to write call notes back. Storman offers a REST API for this kind of connection, and most platforms have something similar or at least an export.

Close (ESC)

Leave write access to gate permissions out of the first version. Once the team has watched the agent handle a few months of calls and trusts its judgement on the easy ones, you can revisit it. Starting narrow costs very little and keeps the security decision where it belongs.

Practical Takeaway

Start by writing down your gate rule: who can authorise access outside office hours, and how that person confirms who they are dealing with. Every other decision in this post follows from that one.

Here is the sequence we would work through. First, pull a month of after-hours calls and sort them into lockouts, payment holds, enquiries and emergencies. Second, write the gate rule and the emergency list, and have the owner sign off on both. Third, set up an on-call roster with a backup and check every number on it. Fourth, find out what your management and access control software can expose, especially whether a code reset can be sent to the tenant's registered mobile.

Those four steps are worth doing with or without an agent. They turn an informal after-hours arrangement into a written process, and a written process is what an agent needs to follow.

With that in place, the build is modest. An agent that answers every call, sorts it, handles enquiries and payment links, and escalates the rest to a person with the context attached. It is the same pattern behind CallCover and our after-hours playbook for property managers, shaped around a gate instead of a front door. You can see the wider range of agent work we do, or get in touch to talk through your site.