Client Experiences
What Organisations Are Saying
Candid accounts from clients across Hong Kong's retail, logistics, media, and professional services sectors.
← Back to HomeOrganisations served
Average satisfaction rating
Years of AI practice in HK
Clients who return for further work
Reviews
Client Testimonials
Chan Lai-ying
Operations Director, Sheung Wan Retail Group
We manage over 4,000 SKUs across three locations. Before the inventory model, our reorder decisions were based on gut feel and spreadsheets. Phosphene took the time to actually understand our supplier relationships and seasonal patterns before building anything. The model has reduced our overstock spend noticeably — and our team actually trusts the recommendations because they can see how they're generated.
February 2026 · Inventory Optimization
Wong Ka-ho
Head of Trust & Safety, HK Marketplace Platform
Content moderation at scale for a bilingual Hong Kong platform is genuinely complex. Phosphene understood that from the start — they didn't try to adapt a generic English-only system. The calibration process was thorough and they were willing to discuss edge cases openly. Our human review team has a much more manageable workload now, and false positives are significantly reduced.
January 2026 · Content Moderation
Priya Lal
Chief Digital Officer, Kowloon Logistics
The Readiness Diagnostic was exactly what we needed before committing to a larger AI investment. The output was candid — they told us our data infrastructure wasn't quite ready for the inventory model we'd been considering, and they gave us a clear path to get there. It saved us from an expensive mistake. We're now working through their recommendations and plan to engage them for the implementation phase.
January 2026 · AI Readiness Diagnostic
Lo Chi-ming
CEO, Mid-range Fashion Retailer, Hong Kong
I'd spoken to two other AI consultancies before Phosphene. Both were convincing but vague. Phosphene was the only team that asked to see our actual data before proposing anything — and then came back with a scope that reflected what we genuinely had to work with. That honesty gave me confidence to proceed. The inventory model has been running for four months and performs as described.
February 2026 · Inventory Optimization
Aisha Rahman
VP Product, Community Platform, Hong Kong
The moderation system has genuinely helped. What I appreciated was Phosphene's willingness to discuss the tension between safety and expression openly — they didn't just maximise for "catches" which would have over-moderated our community. The audit process built into the engagement caught a category of false positives early. Good technical work and thoughtful design.
January 2026 · Content Moderation
Ng Tsz-hin
General Manager, Fintech SME, Central
The Readiness Diagnostic was more useful than I expected. I went in thinking AI was either for us or not — the report showed a much more nuanced picture: some areas well-positioned, others needing work first. The recommendations were specific, not generic. One week, minimal disruption, clear output. Exactly what was described.
January 2026 · AI Readiness Diagnostic
Case Studies
Detailed Client Journeys
Multi-Location Retailer Reduces Overstock by 28% in Four Months
Challenge
A mid-size Hong Kong retailer with 3,500 active SKUs was managing reorders manually. Overstock was tying up capital; stockouts were causing lost sales on fast-moving items. The complexity of their supplier mix — spanning local and Mainland sources with varying lead times — made the problem difficult to manage with spreadsheets.
Solution
Phosphene built an inventory optimisation model trained on 3 years of sales data, supplier lead times, and seasonal patterns including Lunar New Year demand spikes. The model produced SKU-level reorder recommendations reviewed by the operations team before action. Integration with their existing ERP required minimal custom development.
Results
After four months of live operation, the client reported a 28% reduction in overstock holding costs and a 31% reduction in stockout events. Buyer time spent on manual reorder calculations dropped by more than half. The operations team described the recommendation outputs as interpretable and reasonable — an important condition for adoption.
Community Platform Cuts Review Queue by 60% Without Over-Moderating
Challenge
A Hong Kong content platform with a bilingual user base was struggling to maintain content standards without overwhelming their small moderation team. Off-the-shelf moderation tools consistently misclassified Cantonese-English mixed content, either allowing policy violations through or flagging benign posts — both damaging to user experience and team morale.
Solution
Phosphene developed a moderation system trained on the platform's own content and policy definitions, with particular attention to code-switching patterns in Cantonese-English posts. Sensitivity thresholds were calibrated collaboratively with the client's trust and safety team. The human review workflow was redesigned to surface the most ambiguous cases clearly and reduce reviewer fatigue.
Results
Human review volume dropped by 62%. Cases escalated to human review were, on average, genuinely ambiguous — reviewer focus improved substantially. False positive rate fell significantly compared to the platform's previous generic tool. The platform's trust and safety lead described the system as the most contextually accurate moderation tool they'd worked with for HK-specific content.
Get in Touch
Ready to Work Together?
We respond to all enquiries within one business day. Tell us about your situation and we'll take it from there.
Send a MessageCredentials
Professional Standing
HKCS Member
Hong Kong Computer Society
PDPO Compliant
Personal Data (Privacy) Ordinance
ISO 27001 Aligned
Information security management
InnoTech HK Recognition
Responsible AI in Practice, 2023
Add Your Organisation to the List
Join the organisations across Hong Kong who've found a thoughtful, honest AI partner in Phosphene.
Get in Touch