Proof & Validation

Evidence packs, data boundaries, and validation paths you can review before a pilot.

KynticAI stores exact authorised customer data inside the customer-owned data plane, links it through a Rust relationship engine, and serves governed evidence packs to local or open-source LLMs. Cloud remains licensing, updates, support, health, accounts, aggregate usage, and commercial metadata.

Request Technical Walkthrough Review Proof Pack Scope Evidence Pilot

Proof Pack

What is validated, what is scoped, and what stays private.

Public proof should be useful without exposing secrets or pretending pilot-specific work is already finished. These materials are the buyer-safe review map.

ScopedUpdated 2026-06-16

Cloud paid-pilot surface

The private cloud repo contains the operator-assisted commercial surface for first pilot conversations: accounts, contacts, subscriptions, licences, entitlements, data-plane registration, aggregate usage, downloads, support, audit, health, OpenAPI, GraphQL, and lead handling.

Source: Cloud control-plane repo and paid-pilot readiness checklist

Ready2026-05-28

Waitlist and contact path

The public contact path supports product updates, pilot requests, and investor access requests with manual follow-up.

Source: Contact-form safety checks and hosted request flow

Ready2026-05-28 and 2026-06-15

Marketing launch surface

The public site, product pages, scenario pages, ROI journey, waitlist path, and investor materials are in place for first conversations.

Source: IONOS go-live checks and current marketing site state

Scoped2026-05-29 to 2026-06-15

Fortress evidence runtime

The enterprise runtime story shows how a synthetic source can move through local embedding, vector write, and L3 evidence synthesis.

Source: Enterprise runtime notes and local L0-to-L3 benchmark material

Private2026-06-16

Enterprise core engine benchmark

A privacy-safe synthetic, production-shaped proof lane passed for the Enterprise benchmark and pipeline proof: 100,000 vector seed rows, 101,000 final rows, 500 measured vector/search samples, and 19/19 ENT-008 checks.

Source: Private ENT-041/ENT-008 benchmark proof report

Ready2026-06-15

Scout integration surface

Scout provides the open-core path for source registration, selector shaping, snapshots, evidence packs, APIs, and developer-facing integration behaviour.

Source: Scout integration docs and local package validation

Scoped2026-06-15

Importance and Clarity product lanes

Importance Engine supports signal weighting and response pressure; Clarity Engine resolves ambiguous intent before model or agent routing.

Source: Importance Engine and Clarity Engine merged product work

Verified proof run16 June 2026

Core engine proof, shown without exposing private logs.

The Enterprise benchmark lane passed using privacy-safe synthetic, production-shaped data in an isolated proof lane: 100,000 vector seed rows, 101,000 final rows, 500 measured vector/search samples, and 19/19 ENT-008 pipeline checks. The full evidence pack stays in private review.

Boundary: this is technical validation, not a production SLA, customer deployment, regulated compliance claim, or live-customer result.

PASS

ENT-041 benchmark

PASS

ENT-008 pipeline

100k

synthetic seed rows

19/19

pipeline checks

# sanitised verification summary
RUN_ID=20260616T050704Z
PROOF_LANE=ENT-041/ENT-008
DATASET=privacy_safe_synthetic
ENT041_BENCHMARK=PASS
ENT008_PIPELINE_PROOF=PASS
VECTOR_SEED_ROWS=100000
FINAL_VECTOR_ROWS=101000
MEASURED_VECTOR_SEARCH_SAMPLES=500
ENT008_CHECKS=19 passed / 0 failed
RAW_CUSTOMER_DATA=none

Technical Workflow

Source systems -> private data plane -> Rust weighting -> evidence pack -> local AI -> next action.

Source systems

Email, web, registration, CRM, opportunity, support, usage, billing, and outcome systems hold the authorised signals.

Exact private data plane

Exact authorised customer data stays inside the customer-owned data plane with role, purpose, audit, retention, and encryption controls.

Rust relationship engine

The engine links entities, events, recency, contradiction, similar patterns, weights, confidence, and caveats.

Weighted evidence pack

The output is a governed pack containing cited facts, similar successful pattern, confidence, recommendation, and next action.

Local/open-source LLM

Approved local or customer-approved open-source models receive the evidence pack, not raw operational records.

Next-best action

A user sees an evidence-backed recommendation that can be reviewed, challenged, audited, or ignored by an accountable human.

Evidence-Pack Schema Example

A privacy-safe synthetic example backed by a real validation path.

The public schema shape shows what a local/open-source LLM receives: recommendation, similar successful pattern, confidence, next action, caveats, and provenance. It does not require raw customer records to leave the customer-owned data plane.

{
  "schema": "kynticai.evidence_pack.example.v1",
  "privacySafeSyntheticExample": true,
  "validationPath": "public shape backed by local validation path; customer pilots replace this with authorised data",
  "subject": {
    "entityType": "contact",
    "emailAddress": "alex.morgan@example.invalid",
    "crmContactId": "crm_contact_synth_4821"
  },
  "rawDataBoundary": {
    "rawOperationalDataLocation": "customer-owned data plane",
    "cloudControlPlaneReceives": [
      "licensing metadata",
      "update metadata",
      "aggregate usage",
      "health state",
      "support and commercial records"
    ],
    "cloudControlPlaneDoesNotReceive": [
      "connector credentials",
      "raw customer records",
      "prompt payloads",
      "source exports"
    ]
  },
  "authorisation": {
    "purpose": "sales_next_best_action",
    "role": "revenue_operations",
    "auditId": "audit_synth_2026_06_16_001",
    "retentionPolicy": "pilot scoped"
  },
  "sourceSignals": {
    "emailReply": "positive reply received",
    "meetingBooked": true,
    "webSearch": "integration pricing comparison",
    "pricingPageVisit": { "count": 3, "lastSeenHoursAgo": 6 },
    "accountRegistration": "trial created",
    "opportunity": "open expansion opportunity",
    "supportTicket": "resolved API timeout question",
    "productUsage": "usage rising over 7 days",
    "billingStatus": "active",
    "wonLostSaleOutcome": "similar won and lost outcomes available for comparison"
  },
  "relationshipWeights": [
    { "signal": "meetingBooked", "weight": 0.24, "direction": "positive" },
    { "signal": "pricingPageVisit", "weight": 0.19, "direction": "positive" },
    { "signal": "supportTicketResolved", "weight": 0.16, "direction": "positive" },
    { "signal": "billingStatusActive", "weight": 0.08, "direction": "positive" }
  ],
  "recommendation": {
    "summary": "Send ROI proof and schedule the technical sponsor agenda.",
    "valueTarget": "increase conversion probability",
    "similarSuccessfulPattern": "pricing intent + usage rise + support resolved + meeting booked",
    "confidence": 0.78,
    "nextAction": "Send an evidence-backed follow-up that cites the resolved support issue and proposed meeting agenda.",
    "caveats": [
      "synthetic public example",
      "human review required",
      "not an outcome promise"
    ]
  }
}

Validation Paths

Reviewable paths before customer-specific validation.

These paths help technical buyers decide where to start. Customer-specific connector credentials, operational records, and deployment proofs stay in private review.

Ready

Scout SQL source-to-evidence validation

Open-core demo path

Available for technical review

Scoped

Scout n8n local package validation

Workflow automation slice

Available for private review

Scoped

Enterprise connector catalogue review

Private connector families

Scoped by pilot and source system

Planned

Cloud storage document-corpus plan

Document and object-store expansion

Planned for controlled environment validation

Honest Boundaries

Strong proof without unsafe public claims.

Public proof can show

Evidence-pack shape, product workflow, source categories, validation states, static build proof, and privacy-safe synthetic examples.

Private review can show

Source documents, run logs, connector-specific validation, deployment topology, and customer-approved data-plane details under access control.

KynticAI does not claim

Outcome promises, regulated advice, raw operational data in Cloud by default, or complete self-serve SaaS readiness before the commercial operations are live.

Next Step

Choose the review path that matches your risk level.