file · 02 / platform architecturerev · 2026.q3
PLATFORMcore_v4.0 · uaq

One decision engine.
Three operating layers.

The platform ingests roster, demand, SLA and payroll signals from your existing systems, runs them through a per-tenant decision engine, and writes scheduling, dispatch and intervention actions back into the systems your operators already use. No middleware, no parallel reality.

[01]Operating layers
  1. 01

    Forecast

    Site-level demand prediction across hours, channels and skills — replacing spreadsheet rosters and gut-feel staffing curves with calibrated forecasts that close the loop with payroll.

    grain
    15 min · per site · per skill
    horizon
    intra-day · 12 weeks
    closes loop with
    payroll · WFM · ERP
  2. 02

    Scheduling

    Optimization engine producing compliant rosters, routing and dispatch plans in seconds under labor law, contract, skill and fatigue constraints. Re-optimization fires on real-time disruption.

    constraints
    law · contract · skill · fatigue
    re-optimization
    real-time, on disruption
    solver
    deterministic · auditable
  3. 03

    Workforce IQ

    Attrition early-warning, internal-mobility matching and structured pulse feedback. Targets the single most expensive line in a labor budget: unplanned exits.

    targets
    regretted attrition
    signal cadence
    weekly · per cohort
    surface
    named senior lead
[02]Integration surface

The platform sits behind your systems of record. We integrate with HRIS, WFM, payroll, time & attendance, ERP, dispatch and the point-of-sale or telemetry sources that drive demand — through documented connectors, event streams or direct database reads against client-owned infrastructure.

Outbound writes are gated behind explicit policy. Every automated action is logged in an append-only audit ledger that mirrors the client’s own retention requirements. No shadow systems.

systems_of_record
HRIS · WFM · ERP
event_streams
Kafka · webhooks
audit_ledger
append_only
latency_target
< 250 ms
[03]Deployment
  1. phase_01

    Architecture audit

    Structured read of the operating ground: process map, data quality, integration topology. Output is a written assessment, not a pitch deck.

  2. phase_02

    Single-tenant model training

    Models are trained on your data, in an isolated tenant. No cross-customer training corpus, no benchmarking pool, no secondary use.

  3. phase_03

    Shadow deployment

    The engine runs alongside the existing process for a defined period. Decisions are observed and compared before they are wired through.

  4. phase_04

    Live orchestration

    Cut-over to live operation under a named senior lead. Steady-state with quarterly written reviews against the agreed KPIs.

[end_of_file_02]

Reading time is the cheapest part of the engagement. The conversation starts in writing.

Open contact dossier