Lulu Hypermarket

Lulu Retail Holdings

AI Operating System

DeployOne
Viewing as

Operating scope

Country head

Capability

All capabilities

Active handovers

0

Outcomes written back

0

Information density is explain.

Trust and controls

The rules this operating system runs under

An operating system that acts needs limits that are written down, visible in the product, and changeable only in the open. These are those limits.

Six hard rulesEvery approval namedData residency stated

Hard rules

These cannot be switched off inside a product; they change only through the governance cadence.

A human owns every decision that moves money

Agents propose or draft above every stated threshold. Every approval is recorded with a name.

Refusals are first-class

When a human says no, the reason is stored and the threshold can be changed in the open.

Evidence travels with the claim

No number appears without the feed and the moment it came from.

Consent before contact

The consent and suppression register is checked before any call, message or offer.

Modelled values are labelled

Published facts, modelled assumptions and demonstration values are shown separately.

No individual scoring without HR context

People signals are shown to HR with the method attached, never as a bare score.

Decision audit

Every approval, refusal and override is recorded with a name and a reason.

Data residency

Where data is stored and where models run, per country.

CountryData storedModels runNote
UAEIn countryIn regionPrimary lake
Saudi ArabiaIn countryIn countryLocal processing required
OmanIn regionIn regionMirrored nightly
QatarIn regionIn regionMirrored nightly
KuwaitIn regionIn regionMirrored nightly
BahrainIn regionIn regionMirrored nightly

Risks on the table

Stated across all ten verticals, with the level each carries today.

Camera coverage gaps

Medium

17 stores have partial aisle coverage; those gaps are stated, not estimated.

Hypermarkets and stores

Phantom stock

High

Book stock and shelf truth disagree on 2.3% of lines.

Hypermarkets and stores

Expiry capture at source

High

9% of batches are keyed by hand; the agent states its confidence.

Fresh food and production

Donation partner coverage

Medium

Four cities have no collection partner yet.

Fresh food and production

Lead time drift

High

Contract lead times differ from observed by 1.8 days on average.

Supply chain and logistics

Fleet telematics gaps

Medium

Third-party carriers report only at drop level.

Supply chain and logistics

Rider supply volatility

High

Weekend supply swings 22%; promises must stay honest.

E-commerce and delivery

Catalogue quality

Medium

6% of lines lack size or image data for substitution.

E-commerce and delivery

Cost master accuracy

High

Cost changes arrive late for 4% of lines.

Merchandising, pricing and margin

Market observation coverage

Medium

Coverage is thin in two Saudi cities.

Merchandising, pricing and margin

Consent coverage

High

11% of members lack a channel consent record; those are excluded.

Customer, loyalty and marketing

Language coverage

Medium

Malayalam and Tagalog quality is behind Arabic and English.

Customer, loyalty and marketing

Fairness review

High

Screening prompts are reviewed monthly; outcomes are audited by HR.

People and workforce

Skills data

Medium

Skill tags exist for 62% of colleagues.

People and workforce

Meter coverage

High

Sub-metering exists in 61% of sites.

Property, energy and facilities

Asset tagging

Medium

Older sites lack machine-readable asset IDs.

Property, energy and facilities

Statutory boundaries

High

Agents draft; humans post. This is a hard rule per country.

Finance, risk and shared services

Master data

Medium

Vendor duplicates affect 3% of spend.

Finance, risk and shared services

Document quality

Medium

Scanned audits vary; the reader states confidence per field.

Private label and sourcing

Regulatory divergence

High

Six labelling regimes; rules are maintained by humans.

Private label and sourcing