Trust Center

The record of how Propflow has changed.

Safeguards we added, process we tightened, things we corrected, and what is on the roadmap — dated and in plain language. For the full posture, read Security and Data Protection.

Contact routes

Named inboxes with a human on the other side. Include the details below in your first message and we can skip the usual back-and-forth.

security@propflowai.co

Security research + coordinated disclosure

Vulnerability reports, penetration-test findings, and coordinated disclosure from outside researchers.

Include in your first message

  • Reproduction steps, affected URL or endpoint, and rough severity.
  • Your preferred name and whether you want public credit on disclosure.
  • Any PoC artifacts — link or attach; do not include live tenant data.

privacy@propflowai.co

Tenant data, DSRs, and DPA requests

Data subject requests, DPA signature, subprocessor change notifications, and tenant PII questions.

Include in your first message

  • The tenant or account identifier you are asking about.
  • The specific right you are exercising — access, deletion, correction, portability.
  • Your relationship to the account: tenant, operator, or authorized agent.

support@propflowai.co

Operators using Propflow day-to-day

Account issues, Clara behavior questions, work-order disputes, and anything blocking a property team right now.

Include in your first message

  • Property name and the unit or prospect involved, if relevant.
  • What you expected to happen vs. what actually happened.
  • Screenshot or message ID when the issue is visible in the UI.

partnerships@propflowai.co

Press and partnerships

Media inquiries, co-marketing, and integration partnerships. We prefer short first notes over briefing decks.

Include in your first message

  • Outlet or company, and the angle you are working on.
  • Your deadline, if there is one.
  • Whether the request is on or off the record.

The list is in our Privacy Policy. To be notified of subprocessor changes, write to privacy@propflowai.co.

What this page won’t do

Our editorial stance on what belongs here and what does not.

  • We do not publish a cert grid we have not earned. SOC 2, HIPAA, ISO 27001, PCI — not claimed, not implied.

  • We do not quote retention periods we have not reviewed against the running system in the last quarter.

  • We do not list customer audit counts, pentest counts, or logos we have not been cleared to show.

  • We do not answer security questionnaires from a template. Reviewers get responses written against the actual controls.

AI safety & compliance

Four questions Legal and Procurement ask about any AI system — answered by architecture, not by policy documents.

For your Legal / Risk team

Trust is an architecture, not a promise. Open any row.

Procurement asks four questions: what can the AI say, how do you know what it did, what happens when it’s wrong, and where does our data go. Here are the four rows that answer them — built in, not bolted on, not in a separate “AI ethics” deck.

TCPA: STOP and HELP are intercepted before Clara sees themOpt-out and help keywords are answered by fixed code. The AI never gets the chance to mishandle them.

Every inbound SMS is checked for regulated keywords before it reaches Clara. STOP, START, and HELP are answered by a fixed, non-AI handler — the tenant gets the legally required response, the opt-out state is recorded, and Clara is not invoked on that message. Consent is captured and logged at sign-up with IP, user-agent, and timestamp, and every inbound is validated against that record. There is no path where the AI can accidentally reply after an opt-out, because the AI never receives the message.

What happens
Opt-out keywords answered by fixed code, not the AI
Consent record
IP, user-agent, and timestamp captured at sign-up, checked on every inbound
On opt-out
Tenant receives the legally required reply; Clara is not invoked
Audit
Every inbound logged with the intercept verdict before any AI processing
Fair Housing compliantClara can’t discuss protected-class topics on a tenant or prospect thread — blocked before she responds, not audited after.

Before Clara sends any reply on a leasing or tenant thread, the message is screened for protected-class topics — every Fair Housing dimension listed in the spec below, including the state-extended ones. If the screen trips, that reply is not sent. Clara politely tells the tenant she can’t weigh in on that topic and keeps the conversation moving toward what she can help with. This runs on every turn, on every thread, automatically. Compliance isn’t a vibe or a prompt wish — it’s the answer to “what does Clara literally do when a prospect asks about neighborhood demographics.”

When it runs
Before every outbound reply, on every thread
Scope
Leasing and tenant threads alike — same rules in both modes
Protected dimensions
Race, color, national origin · Religion · Sex (including sexual orientation and gender identity) · Familial status (families with children, pregnant persons) · Disability · Age · Marital status · Source of income, including housing vouchers
Jurisdiction
The same check runs on every thread — including the classes that are state- or locally-extended rather than federal, such as source of income
On a trip
Clara declines politely and redirects; the unsafe reply is never sent
Every action Clara takes lands in the audit trailThe conversation history stores not just what Clara said — what she did, in order, with the results she got back.

Every action Clara takes on a thread — every lookup, every decision, every vendor contact, every work-order change — is persisted alongside the human-readable messages. Property managers open any thread in Propflow and replay exactly what Clara did, in order, with the inputs she used and the results she got back. We treat the audit trail as a product surface, not as hidden logs — you can walk a compliance reviewer through Clara’s reasoning turn by turn, in plain language.

What’s captured
Every action, every input, every result — per turn, with timestamps
Scope
Same audit applies to leasing and maintenance threads
Retention
For the life of the customer relationship; deleted on offboarding
Visibility
PM dashboard → conversation detail → expandable Clara audit log
Operator-overridable — you’re still the bossAnything Clara does on your behalf can be paused, reverted, or overridden from the PM dashboard. No lock-in on any AI decision.

Every work order Clara opens, every tour she schedules, every vendor she dispatches, every outbound reply she drafts — all of it lands in the PM dashboard and can be paused, edited, reassigned, or rolled back. Clara isn’t a black box that acts on your property without you. She’s an operator whose every move you can see, approve, and change. Trust an agent the way you’d trust a new hire: let her run, check the work, intervene when your judgment disagrees.

What can be overridden
Every work order, tour, vendor dispatch, and outbound reply
How to intervene
Approve, edit, reassign, or roll back from the conversation or work-order view
On pause
Clara stops acting on that thread; a human takes over without losing context
Audit ties in
Every override is logged alongside Clara’s original action

Commitments we stand behind

Not a badge wall. Four practices that are wired into the product, plus one we are still working toward.

TCPA keywords at the edge

STOP, START, and HELP intercepted before the agent loop runs. Consent captured with IP, user agent, and timestamp.

Fair Housing screening

Every outbound leasing reply is screened for Fair Housing violations before send. Failures block the message.

Full audit trail per message

Every tool call, decision, and outbound reply is logged with full context and exportable on request.

Human-in-the-loop on dispatch

Work orders need operator approval before a vendor is dispatched. Sensitive cases escalate to a named human.

Third-party security review

In progress. Scope set, reviewer selection underway. No attestation claim until the report lands.

Run your portfolio
on autopilot

Go from managing your operations to scaling them.

Book a demo

Hi, I’m Clara. How can I help?