TECHNICAL ROADMAP · FY2027–FY2031

Build the ledger first. Earn autonomy after.

Year 1 follows the architecture's build sequence, Phases 0–5. Every year after expands what the platform does only as fast as observability, evals, and review controls allow.

YEAR 1FY2027PHASES 0–5STATUS: STARTS NOW

Year 1 — the build sequence

Straight from the final architecture. MVP definition of done: upload demand → normalized SKUs → evidence-linked matches → versioned landed-cost quote → approved events on the ledger → lakehouse mirror → dashboards.

Phases on the calendar

FY2027 · GANTT — FISCAL YEAR STARTS JUL 2026
gantt
  dateFormat YYYY-MM-DD
  axisFormat %b '%y
  section Build sequence
  Phase 0 · Data & workflow baseline     :p0, 2026-07-01, 45d
  Phase 1 · Ledger foundation             :crit, p1, 2026-07-21, 71d
  Phase 2 · AI ingestion & normalization :p2, 2026-10-01, 85d
  Phase 3 · Matching & quote engine      :crit, p3, 2026-12-01, 106d
  Phase 4 · Lakehouse & evaluation loop  :p4, 2027-02-01, 73d
  Phase 5 · Order, shipment & finance    :crit, p5, 2027-04-01, 91d
  section Milestones
  Ledger live                             :milestone, m1, 2026-09-30, 0d
  First evidence-backed quote             :milestone, m2, 2027-01-20, 0d
  Quote→payment loop closed               :milestone, m3, 2027-06-25, 0d
    
PHASES OVERLAP BY DESIGN · DARK BARS = REVENUE-CRITICAL PATH (LEDGER → QUOTES → FINANCE)
PHASE 0 · Q1 · DATA & WORKFLOW BASELINE

Decide what truth looks like

Artifact inventory, initial product families, required-attribute ontology, event taxonomy v1, review policy v1, tenant and security model.

ontology v1event taxonomyreview policytenant model
PHASE 1 · Q1 · LEDGER FOUNDATION

The append-only core

Postgres schema for documents, evidence, candidate facts, review tasks, events, projections, and outbox. Ledger command API, hash chain, idempotency controls, basic operator workbench.

ledger command APIhash chainidempotencyworkbench v0
PHASE 2 · Q2 · AI INGESTION & NORMALIZATION

Documents in, candidate facts out

Document classifier, field extractor, product normalizer, evidence-span creation, model run logging, review queues for specs and supplier records.

classifierextractornormalizermodel run log
PHASE 3 · Q2–Q3 · MATCHING & QUOTE ENGINE

The revenue path

Product equivalency agent with attribute-level comparison, match approval workflow, supplier quote extraction, deterministic landed-cost calculator, quote versioning and savings report.

equivalency agentlanded-cost calcquote versionssavings report
PHASE 4 · Q3 · LAKEHOUSE & EVALUATION LOOP

Improve through regression, not vibes

Outbox publisher, Iceberg bronze/silver/gold, model evaluation datasets from review outcomes, reporting dashboards, feedback into prompts and ontology.

iceberg mirroreval datasetsdashboards
PHASE 5 · Q4 · ORDER, SHIPMENT & FINANCE

Close the commercial loop

Customer and supplier PO ledger, shipment milestones, invoice and payment tracking, credit accounts and draws, reconciliation agents for PO / invoice / packing list / BOL / payment.

PO ledgershipment milestonescredit accountsrecon agents
QUALITY GATESTHE NUMBERS THAT GATE AUTONOMY

Auto-approval is earned against these

"Facts auto-approved" means the share of committed ledger events whose candidate facts required zero human touches. The rate only rises while every metric below holds on the regression sets — miss one, and the policy engine rolls thresholds back.

METRIC · MEASURED ON REVIEWED SETSY1 EXITY3 EXITY5 EXIT
Field-extraction accuracy, material fields≥95%≥98%≥99%
Match precision at approval≥97%≥98.5%≥99.5%
Correction rate on auto-approved facts<2%<1%<0.5%
Reconciliation auto-match rate60%80%92%
Quote dispute rate<1.0%<0.5%<0.25%
→ Facts auto-approved40%75%90%
YEARS 2–5FOUR TRACKS

After Year 1: four tracks, one gate

Each track advances a year at a time. The gate for more autonomy is always the same: eval scores and review outcomes, never enthusiasm.

Y2 · FY28
Y3 · FY29
Y4 · FY30
Y5 · FY31
Ledger core
Multi-tenant hardeningRow-level security at scale, tenant-scoped rate limits, SLA-grade projections.
Financial sub-ledger GADouble-entry postings, exposure engine, balancing checks in every commit.
ThroughputOutbox → Redpanda stream; partitioned event store for $1B GMV.
Audit-grade at scaleHash-chain verification jobs, regulator-ready export, 7-year retention.
AI layer
Self-serve ingestionBuyer-run review queues; auto-approval policy engine v1 (→60%).
Equivalency at scaleSubstitution-risk scoring tuned per family; triage agent GA (→75%).
Category machineSchema → ontology → matching pipeline automated; new family in <30 days (→85%).
Agentic operationsException-only humans; reconciliation agents run the back office (→90%).
Data platform
Silver/gold GASupplier scores v1, reorder features, dbt-managed marts.
Intelligence productsCategory benchmarks, savings analytics sold as software.
ML in productionReorder prediction, anomaly detection on shipments and invoices.
The data moat, pricedEquivalency graph + cost history as defensible pricing power.
Fintech
Terms pilotManual underwriting informed by ledger DSO and payment history.
Embedded finance launchNet terms in-quote; credit draws on-ledger; capital partner live.
Underwriting modelsRisk features from gold tables; exposure limits automated.
$18M fintech linePortfolio-level pricing; losses beat bank benchmarks on our data.
NORTH STARAUTO-APPROVAL RATE

One number that proves the whole thesis

The share of facts committed without a human touching them — gated by eval scores, reversible by policy. It is the platform's leverage, the moat's growth rate, and the margin story in one metric.

How the rate is earned

EVALUATION LOOP · FLOW
flowchart TB
  RO["Review outcomes
approve · edit · reject"] --> ED[("Eval datasets
gold tables")] ED --> RG["Regression runs
per model or prompt change"] RG --> PE["Auto-approval
policy engine"] PE -->|thresholds raised| AA["Higher auto-approval rate"] PE -->|scores regress| RB["Thresholds rolled back"] RB --> RO AA --> MV["More volume,
fewer human touches"] MV --> RO
AUTONOMY IS GRANTED BY REGRESSION RESULTS — AND REVOKED THE SAME WAY
Y1 · FY27
40%
Y2 · FY28
60%
Y3 · FY29
75%
Y4 · FY30
85%
Y5 · FY31
90%