Centralization Readiness: Fenix, Lumina, SmartRent Teams — all assume centralized workflows. Clients still site-centric need a centralization roadmap before AI ROI materializes.

Key Takeaways

• The 2025–2026 wave of multifamily AI platforms — Funnel’s Fenix, RealPage’s Lumina, SmartRent Teams — is built on one shared assumption: work is organized centrally, not property by property.
• If your operation is still site-centric — each property with its own leasing inbox, its own maintenance queue, its own way of doing things — these tools will automate your inconsistency, not fix it.
• Centralization is an operating-model change, not a software purchase. Roles, workflows, and data hygiene have to move before the AI produces a return.
• Operators in the 100–2,500 unit range can get centralization-ready in phases: standardize first, pool the work second, then hand routine volume to AI.
• The readiness test is simple: if a task is done differently at every property, an AI agent can’t learn it. Fix that before you sign anything.

What Do Fenix, Lumina, and SmartRent Teams Assume?

All three of the headline AI products in multifamily right now assume the same thing: that work flows to a role or a team, not to a building. That assumption is baked into how each product is designed, which is why centralization readiness — not feature comparison — should be the first question in any evaluation. The evidence is worth looking at directly.

In April 2025, Funnel launched Fenix, a standalone multifamily AI platform built with Sierra, the agent company founded by former Salesforce co-CEO Bret Taylor. Funnel has spent years arguing for renter-centric — not property-centric — operations, and Fenix is that argument turned into software: one AI layer that handles leasing conversations across an entire portfolio, rather than one virtual agent bolted onto each property’s silo.

RealPage’s Lumina is pitched as an agentic “AI workforce” — a set of AI agents that work as a team across leasing, resident services, and operations data. Note the framing: a workforce. Workforces get managed centrally. Nobody hires a workforce and then tells each member to invent their own process at each address.

SmartRent made the assumption explicit. In March 2025 the company expanded its Smart Operations suite with a feature literally named Teams, part of a stated $10 million investment supporting what SmartRent calls the industry’s shift toward centralization. Teams lets work orders be assigned at the group level instead of to an individual at a single property — which only makes sense if you’ve pooled your maintenance techs into a shared-service model across sites.

Three vendors, three different corners of the stack — leasing AI, agentic operations, maintenance and smart devices — and one identical bet. When I evaluate technology for clients, that kind of convergence is the signal I take most seriously. Vendors don’t independently agree on an operating model by accident. They agree because their largest customers already run that way.

Why Do AI Pilots Stall at Site-Centric Operators?

Here’s the pattern I’ve seen repeatedly, and it has nothing to do with the software failing. An operator buys an AI leasing agent or an automated maintenance triage tool. The demo was impressive. Ninety days in, the numbers are flat, the site teams have quietly routed around the tool, and the renewal conversation is awkward.

The root cause is almost always the same: the AI was dropped into five or ten different operations that happen to share a logo. Property A answers leasing emails within an hour and books tours through the CRM. Property B lets the phone ring to voicemail and keeps tour appointments in a paper book. An AI agent trained to hand off qualified leads to “the leasing process” can’t succeed when there are ten leasing processes, none written down.

In practice, three failure modes show up over and over at site-centric operators:

• No consistent handoff point. AI handles the first touch, but every property expects the handoff at a different stage, through a different channel, to a different role. Leads leak at the seam.
• Dirty, fragmented data. Unit statuses, make-ready timelines, and pricing sit in different states of accuracy at each site. The AI confidently quotes wrong availability, and the site team loses trust in it on day three.
• Nobody owns the exceptions. Centralized operators have a person or pod whose job is to handle what the AI escalates. Site-centric operators escalate to “whoever is at the desk,” which at a 150-unit property is often nobody.

None of this is an argument against the tools. It’s an argument about sequencing. When I ran IT at the enterprise level, the projects that failed were almost never bad technology — they were good technology installed on top of a process that wasn’t ready to receive it. Multifamily AI in 2026 is the same story with new vocabulary.

How Do You Know If You’re Centralization-Ready?

You don’t need a consultant’s maturity model to answer this. You need honest answers to a handful of questions. If you answer “no” or “it depends on the property” to most of these, you’re not ready for a portfolio AI layer yet — and buying one will burn budget and staff goodwill.

• Is there one written leasing workflow — from first inquiry to signed lease — that every property follows, with the same stages and the same system of record?
• Can someone who doesn’t work at a given property see its real-time unit availability, pricing, and make-ready status without calling the site?
• Are maintenance requests triaged by the same rules everywhere, or does each maintenance lead decide priority by feel?
• Do you have any role today whose scope crosses properties — a floating leasing specialist, a shared bookkeeper, a regional maintenance tech?
• If a prospect emails two of your properties, do they get a consistent experience, or two different companies?

That last one is the cheapest diagnostic in the business: mystery-shop two of your own properties in the same week and compare the experience. I’ve watched operators discover more about their centralization readiness from that exercise than from any vendor assessment.

One caution for smaller operators: don’t confuse centralization with headcount reduction. At 100–500 units, you may not have enough volume to justify a dedicated central leasing pod, and that’s fine. Centralization at your scale means standardization plus pooling — same process everywhere, and work that can flow to whoever has capacity. The org chart change is optional; the process discipline is not.

A Centralization Roadmap for 100–2,500 Unit Operators

The large REITs spent years and dedicated project teams on centralization. You don’t have that, and you don’t need it. The path below is sized for operators who have a head of operations, site teams, and not much slack. Each phase produces value on its own, even if you never buy an AI product at the end.

Phase 1: Standardize (60–90 days)

Pick the two workflows with the highest volume and the most variation — almost always leasing follow-up and maintenance intake. Document the best version currently running at any of your properties, in plain language, one page each. Then make every property run it. Expect resistance; site teams have usually built their local process for a reason, so harvest those reasons into the standard instead of steamrolling them.

While you’re at it, fix the data those workflows touch. Unit status, availability dates, and work-order categories need to be accurate and current in one system. This is unglamorous work. It is also the single biggest determinant of whether an AI agent will later quote your availability correctly.

Phase 2: Pool the work (90–180 days)

Move one function from “each site does it” to “one place does it for all sites.” The usual first candidates:

• Lead response and tour scheduling — one inbox, one phone tree, one scheduling calendar across the portfolio.
• Maintenance triage and dispatch — requests land in one queue, get prioritized by the same rules, and are assigned to whoever is closest and qualified. This is exactly the model SmartRent’s Teams feature assumes.
• Renewals and delinquency follow-up — calendar-driven, script-driven work that gains consistency immediately when pooled.

Pick one. Run it centrally for a full quarter. Measure response times and conversion before and after, because these numbers become your baseline for judging any AI vendor’s claims later.

Phase 3: Automate the routine layer (180 days and beyond)

Now the AI conversation gets real, because you finally have what the tools need: a consistent workflow to plug into, clean data to draw on, and a central team to catch escalations. Whether you evaluate Fenix for leasing, Lumina for broader operations, SmartRent for maintenance and devices, or a competitor to any of them, the evaluation changes character. You’re no longer asking “will this work here?” — you’re asking “how much of our documented, measured, pooled workflow can this take over, and at what cost per unit?” Pair that evaluation with security diligence: the questions in what to ask a proptech vendor about data security before you sign apply doubly to platforms that will touch every lead and work order in your portfolio.

That’s a question you can actually negotiate on. Operators who show up with baseline metrics and a standardized process get better pilots, better pricing, and honest answers, because the vendor can see their product will get a fair test.

What Does This Mean for AI ROI Timing?

The uncomfortable summary: for a site-centric operator, the realistic timeline to AI ROI is the centralization timeline plus the AI deployment timeline. Skipping the first part doesn’t shorten the total — it usually lengthens it, because a failed pilot costs a budget cycle and, worse, poisons the well with your site teams for the next attempt.

There’s also a portfolio-value angle worth naming. Vendors are pricing and packaging for centralized operators because that’s where their products perform. As agentic platforms mature, the operational gap between centralized and site-centric portfolios compounds: one group gets faster lead response, cheaper maintenance resolution, and leaner staffing per unit, and the other pays the same software prices for a fraction of the result. The roadmap above isn’t just preparation for a purchase. It’s how mid-size operators keep pace with institutions that have full transformation teams.

And if you never buy the AI at all? You still end up with documented workflows, cleaner data, and pooled work — which is to say, a more sellable, more manageable portfolio. That’s the rare technology-adjacent project where the fallback outcome is worth doing on its own.

Frequently Asked Questions

Do I need to centralize before buying any AI tool?

Not every tool. Single-property, single-task tools — an after-hours answering service, a smart-lock system at one asset — work fine site by site. The portfolio-level platforms like Fenix, Lumina, and SmartRent Teams are the ones that assume centralized workflows, because they’re designed to operate across your properties as one system.

Is centralization realistic below 500 units?

Yes, with a narrower definition. At that scale, centralization means one standard process and one shared queue per function, not a dedicated central office. A two-person team covering leasing follow-up for four properties is centralization. The discipline matters more than the org chart.

Won’t centralizing hurt the resident experience at my properties?

It cuts both ways. Residents lose a little “I know Maria at the front desk” familiarity and gain faster response times and consistency. The operators who do this well keep a human presence on site for high-touch moments and centralize the high-volume, low-touch work — which is exactly the split AI platforms are built to serve.

Which function should I centralize first?

Whichever has the clearest metrics and the most pain — for most operators that’s lead response. It’s measurable (speed-to-lead, tour conversion), the volume justifies pooling, and it’s the function where the AI leasing products will later deliver the most visible return.

What to Do This Week

Mystery-shop two of your own properties: submit a lead inquiry to each and log what happens over 72 hours — response time, channel, tone, follow-up cadence. If the two experiences differ meaningfully, you’ve just documented your centralization gap with an afternoon’s effort, and you have the first exhibit for the standardization conversation with your team. Do that before you take the next AI platform demo, not after.

Sources

• PR Newswire — Funnel launches new standalone multifamily AI platform: Fenix, powered by Sierra (April 2025), retrieved 2026-07-10
• RealPage — Lumina AI Workforce for AI Property Management, retrieved 2026-07-10
• Business Wire — SmartRent Expands AI-Powered Smart Operations Suite to Support Centralized Teams (March 2025), retrieved 2026-07-10
• National Multifamily Housing Council — Research & Insights, retrieved 2026-07-10