What clients say about working with us.

The Situation

A Tuen Mun manufacturer was conducting visual inspections manually alongside automated sensor monitoring. The two data streams weren't connected, leading to defects that sensors flagged but visual checks missed — and vice versa.

What We Did

We designed and built a multi-modal system that correlates camera imagery from the production line with real-time sensor readings. The system generates alerts when both streams indicate potential issues simultaneously, and logs patterns for review.

Outcomes

Combined defect detection rate improved by 34% over the preceding six-month average. Manual inspection requirements for standard production runs were reduced significantly, freeing team capacity for higher-complexity checks.

Duration: 11 weeks

The Situation

A compliance function in a mid-sized financial services firm was spending substantial staff time on data gathering and report preparation that involved little substantive judgement. Leadership wanted to understand the automation landscape before making any technology commitments.

What We Did

Over two weeks, we interviewed process owners across the compliance team, observed their working methods directly, and mapped the information flows underlying each activity. We produced a written report ranking eight automation candidates by implementation effort and expected time savings.

Outcomes

The client used the report as the basis for a board-level investment proposal. Three of the recommended automations were approved and are now in development — two internally and one with Ferrotype. The report's analysis directly informed the scope and prioritisation of that investment.

Duration: 2.5 weeks

The Situation

A Causeway Bay retailer had deployed a customer service chatbot that had been well-received at launch. As the product range expanded, the chatbot's ability to answer queries accurately deteriorated — customers were receiving outdated responses or being escalated unnecessarily.

What We Did

We began with a structured technical review of the existing chatbot, then established a monthly maintenance cycle: knowledge base updates aligned with product changes, intent model retraining based on conversation log analysis, and monthly performance reporting with improvement recommendations.

Outcomes

Over four months of maintenance, the chatbot's successful resolution rate — queries handled without human escalation — increased from 54% to 71%. Customer escalation volumes to the live support team fell by roughly one-third.

Ongoing retainer: HK$3,400/month