Method / From operating truth to controlled resolution

Map the chase before automating the message.

Korrd begins with the work as it actually happens: who waits, who decides, what evidence matters, where consent applies, and what must never proceed unattended.

Five gated stages

No live action until the preceding proof exists.

  1. 01
    Audit

    Trace 20 deidentified work orders.

    Map elapsed time, chase events, participant handoffs, evidence gaps, closeout defects, and manager intervention.

    Output: resolution map + baseline
  2. 02
    Configure

    Turn operating judgment into policy.

    Document eligible categories, interrupt conditions, vendors, contact windows, approvals, spending boundaries, evidence, and escalation.

    Output: versioned policy pack
  3. 03
    Shadow

    Read the work without touching it.

    Reconcile Buildium, build case state, draft next actions, and compare Korrd’s proposal with the team’s real decision.

    Output: UAT evidence + corrections
  4. 04
    Approve

    Coordinate through a human gate.

    After all activation gates pass, enable registered channels. A person reviews every outbound message, schedule commitment, approval request, and source-system write.

    Target output: activated L2 pilot
  5. 05
    Prove

    Promote only exact, stable classes.

    Measure edits, safety, identity, consent, delivery, cycle time, and operator effort before one routine template class earns automation.

    Output: keep, change, or stop decision

The 20-work-order audit

Small input. Operationally complete view.

Korrd does not need a data dump to establish fit. The first pass uses deidentified timeline facts and aggregates supplied through an agreed secure process.

01 / SOURCE

What initiated the work?

Category, priority, approval, PMS status, assigned vendor, and known exceptions.

02 / PARTIES

Who owed the next response?

Manager, vendor, resident, owner, or another team—and for how long.

03 / TOUCHES

What did the team have to chase?

Calls, texts, emails, reschedules, status checks, repeated explanations, and corrections.

04 / EVIDENCE

What proved completion?

Photos, notes, invoice, resident confirmation, inspection, or an alternate policy-defined proof.

05 / CONTROL

Which decision needed authority?

Vendor choice, access, estimate, owner approval, spending, dispute, cancellation, or closure.

06 / OUTCOME

Did the PMS reflect reality?

Close date, status, evidence, invoice reference, reopen, and what remained unresolved.

Policy before prompts

The client’s operating rules become versioned system inputs.

Eligibility
Which approved routine categories may enter Korrd, in which market and property types.
Interruptions
Emergency, habitability, accommodation, legal, insurance, access, or dispute signals that pause automation.
Participants
Authorized vendors, client staff, owner approval routes, resident contact rules, and identity boundaries.
Communication
Consent evidence, quiet hours, message classes, reminder limits, language, channel preference, and opt-out handling.
Decisions
Vendor assignment, estimate, spend, access, approval, schedule, cancellation, reopen, and escalation authority.
Closeout
Required photos, notes, invoice references, confirmation, approvals, status mapping, and read-back verification.

Autonomy ladder

Capability advances by evidence, not enthusiasm.

A tenant can pause agents, SMS, email, voice, scheduling, PMS writes, estimates, financial actions, auto-close, or an entire workflow pack independently.

L0

Synthetic

Fictional cases and simulated integrations.

Available now
L1

Shadow

Read current records and compare proposals with real decisions.

Pilot stage
L2

Human approved

Target state after certification: real channels and writes with every consequential action reviewed.

Pilot target
L3

Class limited

Only exact routine actions that satisfy promotion thresholds.

Planned
L4

Routine resolution

Policy-bounded coordination with sampling and kill switches.

Future

Client inputs

Operational facts—not credentials in a form.

Production onboarding happens only after agreement, through scoped secure channels. The public application collects company-level fit information only.

Before the auditDoor count, PMS plan, work-order volume, markets, team structure
Before shadow modeDeidentified examples, policy owners, categories, vendors, evidence standards
Before live messagingConsent basis, approved disclosures, Twilio registration facts, quiet hours, templates
Before writebackRestricted Buildium access, field allowlist, approvers, UAT evidence, rollback owner

Start with operating truth

The first deliverable is clarity, not a login.

Request a deidentified 20-work-order audit to see where resolution time is actually accumulating and whether the founding pilot fits.