CLIENT FEEDBACK
What clients say after working with us
Unedited feedback from organisations that have used our semantic search, retraining, and briefing services.
← Back to Home40+
Engagements completed
96%
Client satisfaction rate
3+
Years serving APAC clients
4.8
Average engagement rating
REVIEWS
Client experiences
Christine Leung
Head of Operations · Insurance sector, HK
We had been frustrated with our internal knowledge base search for years. Staff would give up and ask colleagues directly rather than search. After Lambent implemented the semantic search infrastructure, the number of direct queries to my team dropped noticeably within the first month. The implementation was clean and the handover documentation was thorough.
January 2026 · Semantic Search Infrastructure
Rajiv Menon
Director of Technology · Asset management, HK
Our document classification model had been losing accuracy for about eighteen months and we knew it but kept putting off addressing it. Lambent's retraining service was more structured than I expected — the data validation step alone caught several issues in our training data we were not aware of. The performance evaluation report gave us something concrete to show the business side.
December 2025 · AI Model Retraining
Yvonne Wong
CEO · Professional services firm, HK
I had attended two other AI briefings before this one — both delivered by vendors, both skewed toward selling something. The Lambent briefing was different. Marcus and Sophie had clearly prepared around our sector specifically, and the conversation went where we directed it. The follow-up document was more useful than anything I received from the others. My reservations about AI investment are now more calibrated, not gone.
February 2026 · Executive AI Briefing
Amir Khalil
CTO · Fintech company, Hong Kong
Rajan handled our retraining project with a level of care that larger firms rarely offer. He flagged an edge case in our data pipeline partway through that could have caused problems further down. That sort of attention is hard to find when you are working with people who have twenty active engagements. I have already referred two peers.
January 2026 · AI Model Retraining
Grace Chan
VP Knowledge Management · Law firm, HK
Legal documents have unusual vocabulary and the semantic search needed careful tuning to handle how lawyers actually phrase questions versus how knowledge is recorded. Lambent spent time understanding this before proposing a solution, not after. That made the implementation more effective than I would have predicted from a first engagement.
November 2025 · Semantic Search Infrastructure
Peter Lo
CFO · Mid-sized logistics company, HK
Our board had received three different AI proposals from vendors in six months and nobody could assess them properly. The Executive AI Briefing gave us a common vocabulary and a clear framework for what questions to ask. We ended up declining two of the three proposals with much more confidence than we would otherwise have had.
December 2025 · Executive AI Briefing
CASE STUDIES
Three engagements in detail
CHALLENGE
A mid-sized financial services firm had built a large internal knowledge base over ten years. Staff reported that search was unreliable — relevant documents were routinely missed, while irrelevant results appeared frequently. Adoption of the knowledge base had declined to the point where staff preferred asking colleagues directly.
SOLUTION
Lambent audited the content structure and existing query logs, then implemented a semantic search pipeline using an embedding model suited to financial and regulatory language. Relevance tuning was performed over two iterations using real user queries from the prior quarter. The system was integrated into the existing intranet without changes to the front end.
OUTCOME
Search relevance scores improved by 62% against the baseline evaluation set. Staff-to-staff information requests declined in the first month post-launch. The knowledge management team reported that user adoption had increased within six weeks of go-live.
Timeline: 3 weeks · Service: Semantic Search
CHALLENGE
A technology company had deployed a customer query classification model eighteen months prior. Performance had been acceptable at launch but had since degraded as product lines changed and customer language evolved. The internal data team lacked bandwidth to manage a retraining cycle independently.
SOLUTION
Lambent assessed the current model's performance against a recent holdout set, identifying where accuracy had degraded most significantly. A new training dataset was prepared after validation checks found and corrected labelling inconsistencies. The retrained model was evaluated against the same holdout set and deployed with version control maintained throughout.
OUTCOME
Classification accuracy returned to within 3% of the original launch performance. The labelling issues identified during validation were documented so the client's team could apply checks in future data collection. A retraining schedule was proposed for quarterly cycles going forward.
Timeline: 2 weeks · Service: Model Retraining
CHALLENGE
The senior leadership team of a professional services firm was being asked to approve a significant AI budget. They had limited shared understanding of what AI could and could not realistically achieve, which made consensus on priorities difficult to reach. Two board members were sceptical; others were more optimistic than may have been warranted.
SOLUTION
Lambent conducted an Executive AI Briefing with the six-person leadership team. The session covered the current state of AI in professional services, evaluated one of the proposed internal AI initiatives in framework terms, and addressed the specific concerns and assumptions that emerged from each participant. A follow-up summary document was delivered two days later.
OUTCOME
The team reached a shared position within two weeks of the briefing. One initiative was approved with a narrower scope than originally proposed; a second was deferred pending a clearer problem definition. The CFO described the session as the most useful two hours spent on AI to that point.
Duration: 2 hours + follow-up · Service: Executive Briefing
REACH US
Speak with the team directly
We are a small team and respond personally. If you have a question about whether one of our services is right for your situation, we are happy to have a short call.
CREDENTIALS
Professional affiliations and recognition
HKIAAT Associate Member
Technology auditing certification
HKCS Technology Innovator 2024
Hong Kong Computer Society recognition
PDPO Compliant Operations
Hong Kong Personal Data Privacy Ordinance
Ready to see if we are the right fit?
A brief conversation costs nothing and usually gives both sides enough information to know whether to proceed.
Start a Conversation