For about a year, we were getting six or seven calls a day from people who wanted to talk to their property manager. Broken elevators, lost packages, apartments too hot or too cold. Our website must have ranked next to some management company, because their tenants kept ending up on my line.
It was annoying enough that we got our IT team to build an AI call screener using VAPI, Twilio, and ElevenLabs, mostly to make it stop.
After 20 years and 400-plus Manhattan lease transactions, I know enough about this market to recognize when a technology shift has real implications for office demand.
Initially, I expected a glorified phone tree. But what we got instead was a system that could tell a tenant with a heat complaint from a business owner looking for space, route each one correctly, and hold a conversation that sounded human enough that nobody hung up confused. I tested it hard, tried to break it, and came away comfortable putting it on my own business lines.
So if a boutique commercial real estate brokerage can do that, what happens when a 200-attorney Midtown law firm or a three-floor financial services operation does it at scale?
The investors are already crunching those numbers. A friend, in fact, told me about a call center business that got a $10 million acquisition offer last year and a $2 million offer twelve months later. One anecdote, not a market study, but that kind of repricing doesn’t happen because someone got cold feet. It happens because buyers looked at what AI can handle and decided the headcount behind those phone lines isn’t worth what it used to be.
That same logic is going to run straight through a lot of Manhattan office leases.
The Call Center Just Became Software
When many people hear “AI phone system,” they often picture a slightly better version of “press 1 for billing.” That’s not what I built, and it’s not what’s coming for the commercial real estate market. What changed is that voice AI now handles enough routine inbound volume, at good enough quality, that companies have real reasons to rethink how many people they need in a room wearing headsets.
What a Voice AI System Can Do Right Now
Everyone’s first instinct is, “Oh, so it replaces a receptionist.” Not quite. It replaces the reason you needed ten of them. When software can screen intent, answer the FAQ calls, route correctly, and collect information before a human ever picks up, you stop hiring for volume and start hiring for judgment. That’s a different headcount on a different-sized floor.
The Big Players See It Coming
McKinsey said in 2025 that virtual assistants already handle simple transactional calls and reported a year later that some companies are automating up to 70% of customer contact. Gartner also projected in March 2025 that agentic AI will autonomously resolve 80% of common customer service issues by 2029 and cut operational costs by 30%. What’s more, the BLS also forecasts customer service representative employment will decline 5% from 2024 to 2034 and specifically names automation, self-service systems, and advancing technology as drivers.
After building my own system and watching it handle real calls from real people, none of those projections surprise me. If anything, they feel conservative. That tracks with what I’m seeing — in the past six months, roughly half our initial inquiries have come in after hours, exactly the window where a screened AI call would have caught them.
Where Humans Still Play A Part
Before anyone thinks I’m predicting empty floors across Midtown, let me be clear about what voice AI still handles poorly.
- Complex problems that require judgment don’t fit in a script.
- Emotionally charged conversations that require empathy.
- Regulated processes that carry compliance stakes.
- Exceptions, which by definition don’t follow patterns.
In other words. Voice AI is very good at catching easy, repetitive, high-volume calls that used to require a warm body in a chair. It falls apart the moment a situation requires experience, empathy, or creative problem-solving.
Why This Hits Manhattan Office Space Harder Than You Think
People keep framing AI and office demand around remote work and hybrid schedules. Wrong conversation. Voice AI doesn’t change where people sit. It changes whether certain seats need to exist at all. Once you see it that way, you stop thinking about workplace strategy and start thinking about how much space these tenants actually need.
Phone Volume Used to Dictate the Floor Plan
Call-heavy operations have always been leased the same way. You need X agents to handle Y calls per hour, so you fill a floor with desks, add supervisors, QA staff, trainers, and support managers, and sign a lease big enough to hold all of them. Nobody puts “call center” on the door, but the layout gives it away. Rows of desks. Headsets. The same conversation happening forty times at once. The floor plan followed the phone volume.
Look at Who’s Sitting in Those Seats Right Now
Most of these tenants would never describe themselves as running a call center. But walk past the reception area and into the back half of the floor, and you’ll find departments that run on inbound phone volume every single day.
- Law Firm Intake and Support: Scheduling, case-status questions, document-request routing, and internal help desks within the legal space. A mid-sized litigation firm can have a dozen people doing nothing, but fielding calls from clients who want to know where their case stands.
- Financial Services Operations: Customer service, onboarding support, account inquiries, collections, and service hotlines in financial services. These desks process hundreds of repetitive calls a day, and most of them follow the same script.
- Healthcare and Benefits Administration: Healthcare has appointment scheduling, eligibility checks, billing calls, and referral routing. If you’ve ever called your insurance company and waited on hold for twenty minutes, you’ve talked to one of these floors.
- Property and Facilities Operations: Tenant hotlines, maintenance screening, and dispatch routing. Someone has to answer the phone when the elevator breaks. Right now, that someone sits in a chair on a floor that costs real money.
When AI handles the front end of that volume, the org chart behind it gets smaller. Fewer agents means fewer supervisors, fewer overflow seats, and less need for big back-office footprints.
Those Also Happen to Be Manhattan’s Biggest Lease Signers
Avison Young reports that FIRE tenants lead expansion and new leasing activity, followed by tech and law firms. Those three groups make up over 60% of new and expansion-driven demand. So when staffing models inside those sectors start shrinking, it’s not a footnote. It hits the demand pipeline for half of Midtown.
The part of the office most exposed is not the boardroom, the conference center, or the client floor. It’s the bullpen.
Which Manhattan Office Space Gets Hit First
That bullpen sits somewhere, and Q1 2026 leasing tells you exactly where to look. Manhattan had a great quarter. But the demand was ruthlessly selective about where it showed up.
A Monster Quarter That Played Favorites
Savills reported 12.4 million square feet leased in Q1, the strongest quarter since Q4 2019, with demand concentrated in Class A and trophy buildings. Colliers had it at 11.78 million square feet, the best first quarter since 2014, with availability falling to 13.7% and average asking rents up to $77.55 per square foot.
Midtown drove most of that activity, taking roughly two-thirds of total leasing and posting 6.78 million square feet at an average asking rent of $84.74 with 12.6% availability. Downtown told a different story: 1.04 million square feet, 16.8% availability per Colliers, and CBRE pegged it even softer at 19.3% versus 12.8% for Midtown.
Same city, very different appetite.
The Buildings That Appetite Skipped
Look at where that demand didn’t go, and you’ll see the buildings most exposed to AI-driven headcount cuts. Older Class B. Commodity Class A with nothing to say for itself. Lower-rent admin floors with no amenity story and a value prop that starts and ends with “we can fit a lot of bodies here.” Those buildings were already losing the leasing battle thanks to the flight-to-quality. Now their tenants are about to need half as many bodies, and that value prop evaporates with them.
AI Didn’t Start the Two-Speed Market. It’s Pouring Gas on It.
Manhattan’s office market was already splitting along these lines before anyone put GPT in a phone system. Voice AI just gives the tenants on the weaker side of that split one more reason to shrink or walk. Worth noting, though: Midtown South is not a clean loser here. Colliers reports availability there fell to 13.4% with asking rents at $80.27 per square foot, including record Class B numbers. Buildings with character or a genuine tenant experience are holding fine regardless of asset class. The point is not that every non-trophy building loses. The point is that “cheap seats for warm bodies” stopped being a leasing strategy.
Why None of That Means “Office Is Dead”
Before anyone reads the last few sections and starts panic-listing their Class B holdings, let me walk it back to where the market actually stands. Manhattan office, regardless of how you spin it, still has real momentum.
The Market Is Tightening, Not Collapsing
Colliers shows eight straight quarters of tightening or stable availability and nearly a 30% reduction in sublet space over the last 12 months. CBRE reports Manhattan availability fell 290 basis points year over year to 15.1%, with positive net absorption of 2.06 million square feet. Whatever AI is about to do to certain tenants, it’s doing it to a market that entered 2026 on firm footing.
AI Kills Some Demand, But Creates Other Demand
JLL’s March 2026 research argues AI will pull premium talent toward hubs like New York and increase demand for spaces built around human-AI collaboration, flexibility, and faster business cycles.
That’s already showing up in the leasing data.
AI firms leased roughly 415,000 square feet in Manhattan in Q1 2026, about half of everything they took in all of 2025, and average lease sizes roughly doubled year over year per Bisnow. AI may be shrinking routine-support headcount, but the engineers, founders, product leaders, and deal teams building these tools still want desks near each other.
Law Firms Are Going the Other Direction Entirely
Reuters reported in late 2025 that law firms accounted for 10.5% of U.S. office leasing activity through Q3, double their pre-pandemic share, largely because they’ve committed to in-person work more aggressively than most sectors. AI doesn’t point to less Manhattan office in a straight line. It points to less commodity demand and more premium, collaborative, strategically located demand. Two very different trajectories wearing the same headline.
What Tenants, Landlords, and Brokers Like Me Should Do From Here
Everyone spent the last five years and change arguing about remote work. The next space reduction probably won’t come from where people sit. It’ll come from workflow redesign. That’s a different conversation, and most people in commercial real estate haven’t started having it yet.
If You’re a Tenant
Before you renew or expand, take a hard look at every call-heavy workflow in your operation and figure out which roles generate revenue and build client trust versus which ones exist to answer the same questions all day. Do not lock into a five- or ten-year lease sized around a support headcount that might be half as large in three years. Rethink what the floor looks like, too: fewer identical desks, more escalation rooms, better acoustics, and small rooms built for the conversations that still need a human voice. JLL’s research shows AI is compressing business planning cycles and pushing tenants toward shorter commitments, break clauses, and expansion options. I’d be fighting for that flexibility on every deal right now.
If You’re a Landlord
If your building’s best story is “lower rent and big floor plates,” I’m having a harder time bringing my clients through the door. Some of these buildings honestly may be worth more as something other than office, and the conversion pipeline backs that up. The NYC Comptroller reported in July 2025 that 44 completed and potential conversions totaled 15.2 million gross square feet, could produce about 17,400 apartments, and could absorb over a third of lower-tier occupancy lost since Q4 2019. Cushman reported 2025 conversion starts hit 5.0 million square feet, the highest in 20 years, with a 9.8 million square foot forward pipeline. AI pressure on top of a healthier conversion market means marginal office stock has nowhere left to hide.
What We’re Doing Differently
At Metro Manhattan Office Space, we’ve started weaving workflow analysis into every space conversation we have. We ask clients which teams genuinely need to sit near each other and which ones share a floor only because that’s how it was set up ten years ago.
For me personally, every deal now falls into one of three buckets: space that grows because AI increases collaboration needs, space that shrinks because AI absorbs repetitive volume, and space that should leave the office inventory entirely.
Where This Leaves Manhattan Office Demand
I didn’t set out to write a market forecast after building an AI phone system to stop getting calls about other people’s elevators. But once I watched that system handle real conversations well enough to trust on my own business lines, I couldn’t stop thinking about what it means for the tenants I work with and the buildings they sit in.
Manhattan office demand has genuine momentum right now. The buildings designed around collaboration, talent, and client-facing work are going to keep filling up, because AI makes those environments more valuable, not less. What gets harder to justify are the big, undifferentiated floors that exist because someone needed fifty people answering phones in the same room. That rationale is dissolving faster than most landlords want to admit.
A whole lot of Manhattan office space was built around the assumption that certain jobs required a warm body in a chair. Software can do those jobs now. It can do them well. And that changes the equation on more leases than anyone in this market is comfortable talking about.
