Landmark — Technology Review

Technology overview

Property management at scale — and the technology layer we built to run it.

SubjectThe Landmark Platform FromTroy Muljat · CEO StatusConfidential · Not for distribution

Who we are — core purpose

Opening doors. Building wealth.

That's our core purpose — not a tagline, a promise. Landmark Real Estate Management, Bellingham, Washington: ~4,000 doors, ~10,000 residents, 60+ team members. Spokane opens in 2026, Boise and Vancouver follow — on the way to 20,000 units across six markets by 2035.

Simplify·Standardize·Scale

01

Maintain Accountability

We do what we say we'll do. Every time.

02

Problem Solvers

We don't wait for an answer. We go find one.

03

Drive Change

We lean in where others see disruption.

04

Serve Others First

Owners, residents, teammates — help first.

05

Do What's Right

Even when it's hard. Even when no one's watching.

We run on EOS — rocks, scorecards, L10s. The discipline is "simplify before you multiply" — and when no software on the shelf could deliver that simplicity, we built our own.

What we're doing

A property company that
builds its own software.

Landmark runs a full property management operation — and underneath it, a technology layer we built ourselves: an agentic back office, running on the live book, that does the repetitive operational work. That layer is what this overview is about.

The operation

Landmark Real Estate Management

4,100 residential units · 920K SF commercial
Established brand, ~10,000 residents
The live portfolio the platform was built on

The technology

The Landmark platform

AI agents that code bills, reconcile, and draft statements
Money-safe, audited, human-in-the-loop
In production today — not a roadmap

The market window

AI in property management just
crossed the chasm. The window is open.

78%

Of operators say they've already lost business to AI-enabled competitors

34%

Of property managers using AI — up from 21% in a single year

18–36mo

The window before incumbents close the agentic gap

Traditional operators are already being squeezed — average PM profit margins fell from 22% to 19.2% in two years. Operators that own this capability set the cost structure the rest of the market has to chase. Landmark already owns it.

Sources: EliseAI, The 2025 State of AI in Multifamily (280 operator executives) · AppFolio 2025 PM Benchmark Report (2,000+ PMs) · Crane 2026 Industry Report (100 PM operators)

Where the field is — and where it isn't

Front-office AI is a crowded shelf.
The money-safe back office is empty.

The incumbents — AppFolio, Buildium, Yardi — were architected pre-AI and bolt agents on, gated to premium tiers. The AI-native challengers — EliseAI, Funnel — do leasing chat and CRM, and none of them touch money. An AI that codes a bill is a weekend hack. An AI that cannot double-pay, cannot skip a deposit rule, and logs every decision to a seven-year standard is a year of architecture. We did that year.

Crowded — front office

Leasing chat, CRM, scheduling

EliseAI, Funnel, Knock, ELI+ — billions in valuation
All sit on top of someone else's ledger
None of them move a dollar

Empty — the seam we hold

Agentic back office

Bill coding, reconciliation, owner statements, deposits
Money-safe, audited, human-gated by design
Running in production on a real book today

Why we built it — part one

Before we built anything,
we audited ourselves.

March 2026: a two-day, on-site operational audit of Landmark's own shop — twelve staff interviews from the front desk to the CFO, every system in the stack reviewed, every workflow walked end to end. Not a vendor's pitch deck — a hard look at our own operation.

157

Distinct findings documented across the operation

$411K/yr

Software spend across 16+ tools — before payroll touches a single task

<50%

Of the core PMS's features actually used — confirmed by multiple staff

$445K–$742K in annual savings identified — nearly all of it locked behind what the tools couldn't do.

Why we built it — part two

Every tool we bought hit
the same ceiling.

Read-onlyThe workflow layer couldn't write back. It read the PMS but every action had to be re-entered by hand — staff logging the same data three to five times. $72K a year for what leadership called "zero value, actually negative value."

No edge casesEvery exception broke the automation. Backdated entries, lease breaks, large group applications — each one fell back to a human and a spreadsheet. The result: five shadow spreadsheets, three CRMs, 500–600 receipts hand-keyed every month.

Bolt-on AIThe incumbents' AI was bolted on, gated, and not trusted. Premium-tier paywalls on pre-AI architecture — the people doing the work turned it off.

The verdictThe audit's conclusion: growth to 16,000 units on the current stack is not feasible. No amount of additional subscriptions fixes that.

So we stopped renting other people's software and built the layer ourselves. That platform is ChiefAIO — and the audit is its product spec.

What we built

One system for the
whole operation.

We didn't buy a system and bend our process around it — we built the system around the way Landmark actually runs. Every part of the operation in one place, with AI agents working across all of it.

Leasing

Vacancies, showings, leases, renewals.

Rental Applications

Intake, screening, decisions.

Properties

Units, buildings, the full portfolio.

People

Owners, tenants, vendors, directory.

Accounting

Financials, bills/AP, deposits, receipts.

Proven first

Maintenance

Work orders, AI dispatch, vendor follow-up.

Communications

Messages, documents, letters.

Reporting

Owner reports, scorecards, team KPIs.

Built and running on the live portfolio today — not a prototype, not a roadmap.

Proof — already live

This runs today on Landmark's
real book of business.

343K+

AP bills under management

194K

Invoice PDFs captured & processed

Live

Bills · CRM · work orders · reconciliation

Built end-to-end on Landmark — a working property-management operation and Troy's own portfolio. The platform is real, the data is real, the agents run against it. In two minutes you'll watch it work.

The product — beta, live at leasewa.com

Not a mockup. The live beta.

Landmark platform dashboard — executive overview with portfolio KPIs

1,100+

Agent runs to date — every one logged & replayable

4,808

Units in the platform — the live book, not seed data

End‑to‑end

Leasing, accounting, maintenance, reporting — one platform

Executive KPIs · scorecards · AI receptionist · AI dispatcher · bills & AP — one login instead of sixteen tools.

The AI workforce — live

Every agent is a worker, a supervisor,
and a human who can say no.

Landmark AI agents roster — worker and supervisor pairs per pillar, running on Claude Opus

9M+

Tokens of reasoning against the live book — logged & replayable

270+

Actions held for a human to approve or reject

Opus

Every pillar paired worker + supervisor on Claude's top model

Five pillars, each a worker that does the task and a supervisor that checks it — and anything material waits for a person.

Inside the platform

Not a dashboard — the whole workflow.

Human review queue

Escalations board — actions awaiting human approval across the agents

Every judgment call the agents can't make alone — queued with their reasoning, one click to approve or reject. (Applicant details blurred.)

Plugs into the stack

Integrations — Zillow, CoStar, Craigslist, QuickBooks, screening, payments

Zillow, Apartments.com / CoStar, Craigslist, QuickBooks, screening, payments — syndication and connections built in.

The agents do the work, a person stays in control, and the platform connects to everything around it.

The bigger picture

Four systems, one company — all ours.

The property platform is one piece. We're building the software for the whole company — how we run the business, manage property, win deals, and find the next one — one connected operating picture, not a stack of subscriptions.

Run the business

LANDeOS — Landmark Operating System (EOS)

LANDeOS — EOS: rocks, scorecard, L10s

Manage the property

The Landmark property platform — leasewa.com

The PM platform — agentic back office, 4,800+ doors

Win the deals

Muljat Group Commercial CRM — creproperty.com

Muljat CRM — commercial pipeline & comps

Find the next deal

Whatcom Multifamily Database — business development prospecting

Multifamily DB — every property + its manager, any market

Same team, same standard, same data layer — the first time everything Landmark does lives under one roof.

What makes it ours

One platform. Our agents. Any model.

01 — Under one roof

Everything in one place

Run the business, manage property, do deals — one login, one data model, no integrations to babysit. We own the whole stack instead of renting sixteen tools.

No stitched-together stack

02 — Agentic by design

Bots do the work

Agentic AI handles 25–80% of the work in each function — the repetitive load, done and checked, with a human on anything material.

25–80% of the busywork, gone

03 — Model-agnostic

Any frontier model

Claude today; GPT, Grok, or whatever's best tomorrow — swapped in without rebuilding. We're not betting the company on one vendor.

Future-proof by design

Built in Rust

The language used for systems that can't fail — memory-safe, fast, and correct under load. The right foundation when the software is moving real money every day.

Live demo

Don't take the slide's word.
Watch it do the work.

What you're about to see — three moments

01 — IT WORKS

A real bill, coded live

Pull an actual invoice from the live system. The agent reads it, codes the GL, proposes the action — on real data, not a sandbox.

02 — IT'S SAFE

The guardrail stops it

We trigger an unsafe action on purpose. Watch the deterministic gate block the commit and route it to a human. This is the moment that matters.

03 — IT'S PROVABLE

The full audit trail

Open the transcript. Who, what, why, when — every decision logged, replayable, to the standard an auditor would ask about.

[ Presenter: switch to leasewa.com now. Return here when done. ]

How the work gets done

Agents that do the work —
not a chatbot bolted onto a PMS.

WorkerDoes the task. Reads the data and works out the answer — the GL code, the payment match, the disposition — then proposes it. Models are tiered so cost stays flat at volume: lightweight models triage, the working model does the task, the most capable model is reserved for high-stakes review.

SupervisorReviews the work. A second pass for coherence and completeness — catches what doesn't follow from the data, writes the audit note, decides what a person should see. The hard safety guarantees don't rest on it; those live in code.

HumanIn the loop by configuration. Dollar limits, new vendors, unmapped accounts route to the operations team. Nothing irreversible commits without a person.

The routine runs on its own. Anything irreversible or material is gated — that's what makes it deployable at portfolio scale.

Why this runs in production with real money

The safety isn't the model being smart.
It's the model being unable to skip the check.

Every money-moving or irreversible action passes through a deterministic gate before it commits — code engineers and counsel can read, that no model, however confident, can argue its way around.

Live readRe-checks reality at the moment of commit. Microseconds before any write, the system re-reads the live record. A bill paid out-of-band five minutes ago cannot be paid twice.

Hard rulesDeterministic, version-controlled, lawyer-reviewable. Deposit returns, new vendors, unmapped accounts, over-threshold amounts always stop. Code — not a prompt the model can be talked out of.

Human gateEvery irreversible action needs a person. No confidence score overrides the human approval.

Full auditWho, what, why, when — retained seven years. Every decision with the agent's reasoning verbatim, replayable, to the HUD standard.

Architecture

Built once. Runs on top of
whatever the portfolio runs.

A normalized adapter layer means the agents don't care what the underlying PM system is. Acquire a book on AppFolio, Buildium, or Yardi — the same agent layer drops on top without rebuilding anything.

Agent layer — the 5 pillarsPMS-agnostic by design

▲ ▼

Normalized data model + adaptersone common shape

▲ ▼

AppFolio — live today Next acquired book — scoped adapter Buildium Yardi

AppFolio runs in production now. Each new adapter is dedicated, scoped work — which is what lets the platform travel to every market we enter, not just this one.

What the platform changes

Three things it changes
day to day.

01 — Cost

Lower cost per door

The industry averages 48 doors per employee; the NARPM benchmark is 59. Landmark already runs 60+ — and the agents absorb new doors without new headcount, the inverse of the traditional PM cost curve.

Compounds with scale

02 — Reach

Built to expand

The average operator added just 108 doors in all of 2025 — growing the traditional way is slow. The adapter layer drops onto whatever a new office or market already runs, so the same back office travels with us to Spokane, Boise and beyond.

Travels to new markets

03 — Service

Less busywork, better service

The agents carry the repetitive operational load — bill coding, reconciliation, statements, work-order triage — so the team spends its time on owners and residents instead of data entry.

More time on people

Built for one reason — to manage more doors, in more places, without losing the standard.
Industry data: Crane 2026 survey (100 operators) · NARPM / ProfitCoach 2024 PM Operations Standards

Where we are — and what's next

The rollout, de-risked.

Now → end of July

Testing & tuning

Refining the UI and adding the last components, live.

August

Full beta

Every team in the system, end to end.

September 1

Preliminary go-live

Running in parallel with AppFolio — nothing turned off.

Oct 1 – Dec 1

Official go-live

The window to fully switch over and retire AppFolio.

We run both systems side by side until the new one has earned the switch — live, but never a leap of faith.

The short version

We manage doors —
and build the tools to do it.

A property management company that got tired of the software on the shelf and built its own: agentic, money-safe, live today, and made to scale with us. Happy to show it working anytime.

Troy Muljat · Landmark Real Estate Management visitlandmark.com · the Landmark platform, live at leasewa.com