Legal & trust

AI & Co-Manager Transparency

The X Co-Manager is an owner-facing simulated intelligence lab. Its current audited analysis and explanations use fixed synthetic data and deterministic rules. The AI product description is not evidence of a live language model, autonomous supplier purchasing or proven financial improvement.

Draft for review — not approved or effective terms

These documents are review drafts. They require founder confirmation and independent legal review before adoption. No acceptance of new terms is requested or recorded.

Review version:
phase-3.4.1-draft-1
Prepared:
October 24, 2026
Effective date: not set — awaiting approval

1. Data supporting analysis

The base lab uses the versioned synthetic retail dataset retail-lab-v1: two example stores, six products and sixty days ending 24 October 2026. Trend-radar examples are explicitly synthetic. Import demonstrations accept only approved retail-import-v1 CSV fixtures, not real merchant uploads or live external signals.

Each recommendation is linked to calculation inputs, rule/version, time window and quality evidence. Missing costs, stale inventory, incomplete history, mismatched payments, stockouts and expiry risks can limit or withhold advice. Synthetic evidence is not a validated production dataset and its freshness does not advance automatically with today’s date.

2. Calculations versus explanations

Sales, reconciliation, known-cost coverage, reorder estimates and comparisons come from documented deterministic calculations, not a language model estimating financial figures. Conversation answers use rule-based intent handling and saved evidence. Advisory styles may change presentation or challenge a hypothesis, never underlying facts.

No external AI/model request was found in the current runtime conversation, proactive or import paths. Assets created during development are separate from customer requests. Future external model explanations require provider, payload, retention, transfer, training-use and human-review disclosures before activation; no blanket future promise that customer data never leaves the platform is made.

3. Uncertainty and human control

Forecasts and suggested experiments are limited hypotheses, not guarantees of sales, profit or causal uplift. Historical averages, simple assumptions and synthetic confounders do not replace a prospective business evaluation. Financial scenarios are illustrative, not actual customer KPIs or validated commercial results.

Owners decide whether to review, accept, reject, simulate implementation or evaluate recommendations. Autonomous supplier purchasing, payment execution, live POS/supplier connectors and delivered alerts are not operational. No recommendation may initiate physical enforcement or restrict emergency exit; Guardian uncertainty requires human review.

4. Saved conversations and evidence

Your questions, rule-derived responses/evidence, recommendation snapshots, reasons, lifecycle stages, evaluations and preferences are stored in owner-scoped Cloud records. These records survive reloads; they are not anonymous simply because the sales data is synthetic. Avoid entering sensitive personal data or third-party trade secrets in free text.

Service-only writes derive evidence after authenticated owner reads; clients cannot rewrite approved evidence or access another owner’s history. Founder access does not by itself reveal others’ private conversation/import records. No cross-merchant benchmarking, data sharing or model-training pipeline is enabled. Retention/erasure operations still require approval and dedicated implementation.

5. Report inaccuracies and future changes

Report an incorrect recommendation using the workspace’s product feedback or the privacy/legal contact. Include the relevant rule/scenario and why the evidence appears wrong, without passwords, tokens or unnecessary personal information. Feedback is stored and visible to authorized staff; it is not proof of model training or automatic correction.

Before introducing external AI or real business data, the operator must approve evidence quality, merchant authorization, provider processing terms, lawful bases, confidentiality controls, evaluation and updated bilingual disclosures. No such capability is activated by this draft, and independent legal/retail/language review remains outstanding.

Legal & privacy contact

Nikolov.n.ivaylo@gmail.com
Operating legal entity
Founder confirmation pending
Company registration number
Founder confirmation pending
VAT number, if applicable
Founder confirmation pending
Registered office
Founder confirmation pending
Applicable jurisdiction
Founder confirmation pending
Data Protection Officer, if required and appointed
Founder confirmation pending

Legal review required

Operator identity, lawful bases, retention, transfers, eligibility, liability, jurisdiction and the Bulgarian legal wording require qualified legal approval. This draft is not a compliance certification.