Operating Labs

Where workflows are tested before claims are made.

Our Operating Labs are internal operating environments used for active testing, operational validation and continuous optimisation. They help us understand how systems behave under real workflow constraints. They are not a substitute for authorised client results.

AI Commerce Lab

An internal operating environment for testing AI-powered product research, content production, listing automation and commerce workflows.

45 min → 6 min

first-draft time per SKU

78%

passed review first-pass

On an internal test set of 1,200 SKUs, first-draft listing production time fell from ~45 minutes to under 6 minutes per SKU (Q2 2026 test cycle).
78% of generated multilingual listing drafts passed human review without structural rework (Q2 2026 test cycle).

Internal lab evidence · Not client results · Scope and period as stated

Supports the AI Commerce System

Market Intelligence Lab

An internal intelligence environment for testing how fragmented market signals can be transformed into structured commercial decisions.

12 → 90+

sources monitored per analyst

~2 days → 3 h

report assembly time

Weekly monitoring coverage expanded from 12 to 90+ authorised sources per analyst without added headcount in internal testing (Q2 2026 test cycle).
Report assembly time reduced from ~2 days to under 3 hours, with every material finding linked to preserved evidence (Q2 2026 test cycle).

Internal lab evidence · Not client results · Scope and period as stated

Supports the AI Market Intelligence System

Content Operations Lab

An internal production environment for testing scalable image, video and multilingual content workflows, supporting the AI Content Operations System.

500

brand-constrained images in one working day

82%

first-pass editorial acceptance

Batch production of 500 product images with brand constraints completed in one working day in internal testing (Q2 2026 test cycle).
First-pass editorial acceptance for multilingual adaptations (EN / zh-Hant / DE) reached 82% (Q2 2026 test cycle).

Internal lab evidence · Not client results · Scope and period as stated

Supports the AI Content Operations System

Private AI Lab

An internal infrastructure environment for testing private models, model routing, workflow orchestration and local AI operations.

model families routed under load

10,000

documents in retrieval corpus

Multi-model routing across 3 model families sustained under continuous internal workloads with per-task cost tracking (Q2 2026 test cycle).
Private knowledge retrieval evaluated on a 10,000-document internal corpus with permission-scoped access checks (Q2 2026 test cycle).

Internal lab evidence · Not client results · Scope and period as stated

Supports the Private AI Platform

Sales workflows are currently validated inside client pilots rather than a standing lab.

Measurement

What we measure

Evidence is recorded with its scope and limits.

Output quality
Cycle time
Human intervention
Exception rates
Operating cost
Reliability
Adoption

Lab observations last updated: 2026-07

From lab to client pilot

Lab methods become reusable components only after review. A client pilot still starts with the client's workflow, data, controls and success criteria.

Disclosure policy

Lab observations are always labelled as internal evidence with their scope and time frame. They are never presented as client results. Client names, data and outcomes are published only with written authorisation.

Explore a focused pilot.

If a workflow in your business deserves testing under real conditions, we can define the scope, baseline and decision criteria together.

Explore a Focused Pilot