Phone calls.
The never-ending ring. The directory trees. Begging a robot, "Speak to a human."
Most businesses you call today are running some kind of AI on the other end of the line. Reception was one of the first business functions AI came for, and it came fast.
Why?
Because the job on the other end of a business line is acknowledgment plus accuracy, not connection. Greet, identify, help, route. That's a narrow enough lane for a machine to drive well, and AI got good at patterns before it got good at anything else.
We've done this before. The history of the phone is mostly a history of who got cut out of the call.
A short history of who used to answer
Every communication technology we've built was built to overcome the limitations of the one before it. And every one of them eventually erased the person who used to sit in the middle.
The telegraph operator. Before phones, the closest thing we had to instant communication was the telegraph, and you couldn't use one. Only trained operators could, only from telegraph offices, only in Morse. Western Union peaked in 1929 at over 200 million telegrams a year, employed tens of thousands of operators, and ran an entire urban workforce of delivery boys. The last Western Union telegram was sent on January 27, 2006. An entire profession, gone, swept aside by a tool the public could finally use without permission.
The switchboard operator. When Bell, Meucci, and Gray (history lets you pick a side) introduced the telephone, they didn't eliminate the human in the middle. They just moved her. At its peak in 1947, AT&T employed over 350,000 switchboard operators, almost all women, almost all answering "Number, please?" and physically plugging cables between sockets. Direct dial rolled out between 1919 and the 1960s and erased roughly 90% of that workforce. You started dialing the number yourself.
The mail clerk and the executive assistant. Ray Tomlinson built email in 1971 so people sharing a mainframe could leave each other messages. He picked the @ sign because no human's name had it. Email grew into the thing that finally killed telephone tag, and along the way it killed the office mail clerk, the inter-office memo, and roughly half of the executive assistant's job description.
Three tools, three job categories gone.
The pattern is consistent. Every new way to talk replaced someone who used to listen.
The offshoring decade
By the late 2010s, we'd more or less iterated to the end of the tools. So we started moving the labor instead.
Between roughly 2018 and 2024, you could call a property management office in Knoxville (go Vols) and have the call answered by someone in Manila, Bogotá, or Cape Town. The Indian and Filipino BPO industries each grew past $30B a year, employing a combined three million people. The smart operators ran a follow-the-sun model: 24-hour coverage built entirely out of timezone arbitrage. It worked. For plenty of PM teams, it still works.
This week, one of the OGs of outsourced reception in our industry, Virtually Incredible's Phone Tenders, announced it's sunsetting that service at the end of the month. Virtually Incredible itself stays open and continues to offer trained virtual assistants. The piece that's going away is the always-on, software-routed call center.
Why?
Per founder Todd Breen, the legacy software platform their call center runs on is being shut down by its vendor, and there isn't a comparable replacement on the market at a price point that works for their customer base. So in this specific case, it wasn't AI that closed the door. It was the rails.
The reaction was immediate. Across the comment thread, VI's customers were openly asking what to do next. Some were looking for another outsourcer. Some were looking at AI. The conversation that used to happen privately, between PM owners at a conference, happened in public in a single afternoon. The market is repricing in real time, and the customer base is split on which way to go.
But the second-order force is the same one tugging on every option in the market. Three years ago, voice AI cost about fifty cents a minute, and it sounded like it. Today, production-quality voice AI runs closer to five to ten cents a minute, and most callers can't tell anymore. A US receptionist is twenty to thirty dollars an hour fully loaded. A Philippines-based VA is eight to fifteen. AI is cents per call. Below a certain price, the floor moves. Every other option, AI included, has to decide whether it can live below the new floor or whether it has to sit above it for reasons that aren't price.
So is AI actually better than a human on the phone?
Sometimes. Not always. But the question itself is the wrong frame. The real question is which job, exactly.
AI is coming for reception in particular because reception is the most patterned communication work in a business.
Most inbound calls fall into a handful of buckets:
Someone wants to give you money (a prospect, a renewal, a payment).
Someone wants you to fix something (a leak, a lock, a busted unit).
Someone wants to know something a website should have already told them.
Someone wants to sell you SEO services (again).
Two of those four are worth answering. One eats forty percent of the day. And the last…we all wish would just go away to begin with.
A great human receptionist can triage that. A great human receptionist also costs $45–65K a year, works one shift, takes lunch, takes vacation, and is allowed to have a bad Monday. A well-built system doesn't.
This is not AI coming for every phone call. Pet adoption hotlines, grief counselors, the friend who needs you to pick up on a hard night, none of those are getting AI'd anytime soon, and they shouldn't.
The places where AI on the phone works are exactly the places where the job is acknowledgment plus accuracy, not connection. Front-desk triage for a 1,200-unit portfolio. Maintenance dispatch at 11 PM on a Sunday. Owner-relations inbound during the dinner hour. The work where the caller doesn't want intimacy. They want their problem moved one step closer to solved.
So if reception is the narrow lane where this works, the only real question left is whether the system in the lane actually does the job.
Which is where I get to talk about ours.
Here's one of our AI Workers, our Maintenance AI, handling a real tenant call. Same voice family as the Receptionist, same dispatch logic, different specialty.
Better than a great human receptionist on her best day, with full context, no other calls in queue, and every property in your portfolio memorized?
No.
Better than the call that didn't get picked up because she was already on another line, or out sick, or it was 7:42 PM, or it was a Sunday, or it was the fifth time that hour someone called about the same broken HVAC?
Yes.
That's the bet. Not human versus AI.
Picked up versus missed.
What your team actually gets
Your team stops triaging. The Receptionist sorts calls at the door, by intent. Leasing inquiries go to leasing. Maintenance emergencies go down the on-call path you've defined. Owner questions go to the right portfolio manager. The SEO pitch never makes it to a human.
Your team gets their evenings back. If you've told us leasing works 9 to 6, a 7:42 PM call doesn't ring an empty desk, and it doesn't ring a tired person at home. It gets handled, and a structured message lands in the right inbox for Monday morning, already prioritized.
The calls your people do pick up are warmer. Three or four qualifying questions get answered before any human says hello, and the transfer comes through warm to the right extension. The leasing agent who takes the call already knows the prospect's name, the unit, the move-in window, the budget. The maintenance coordinator already knows the address, the issue, whether it's an after-hours emergency or it can wait. The first ten seconds of every call stops being "let me get your name."
Your knowledge base finally does the answering, not the guessing. When the Receptionist handles a policy question, it pulls from your own documents, with confidence thresholds so it knows when to hand off instead of invent.
More on the mechanics in this week's Feature Spotlight
Speak to a human, for real this time
The phrase at the top of this piece is what we say when we're stuck in a phone tree. Speak to a human. What we actually mean is: connect me to the person who can fix this. We don't want a human in the abstract. We want the right one.
For 150 years the people who connected us to the right human kept moving farther away. From the switchboard girl at the local exchange, to the answering service in the next town, to the BPO agent in another hemisphere. Each generation built a tool that put the listener somewhere new. This one puts it everywhere at once, and grades itself against the only standard that ever mattered.
Talk to people. Pick up when they call. If you can't, build something that does it the way you would.
Or come hear what we built. → Book a demo
THE BUILD LOG
A few highlights from the last two weeks across the Haven codebase:
NEW: Self-Showings, on the hardware you already use. Haven now runs self-guided tours through Codebox lockboxes and any smart lock supported by Seam (Igloo, Schlage, August, Yale, and the rest). Prospects book a tour, get verified, and receive a one-time access code for the window you've approved. No leasing agent required, no after-hours showings on your team's calendar. Every visit, every code issued, and every entry rolls up into a single activity view.
More coming soon…be on the lookout!
Madeline