What it's actually like to work with Voxelith
Feedback from organisations we've worked with — shared as candidly as they gave it to us.
Back to HomeClients served
Average rating
Satisfaction rate
Return clients
From recent engagements
William Leung
Head of Research, Asset Management Firm
The knowledge graph Voxelith built has genuinely changed how our analysts find connections across our research archive. We went from doing manual cross-referencing — which nobody had time for — to running structured queries that surface relationships automatically. The delivery was thorough and the documentation actually made sense.
Knowledge Graph Construction — January 2026
Patricia Chan
Legal Operations Manager, Mid-size Law Firm
We engaged Voxelith to extract key clause data from a large volume of legacy contracts. What impressed me was how carefully they defined the extraction schema with us before writing a single line of code. The human review queue they built in was essential — we're dealing with material that we can't afford to get wrong, and they understood that from the start.
Text Mining and Extraction — December 2025
Henry Tang
Director of Operations, Financial Services Firm
The AI Readiness Training was more grounded than I expected. Our team had been using AI tools somewhat haphazardly, and the sessions helped them develop a more consistent approach to evaluating what these tools produce. The exercises were practical, not theoretical. One note: session pacing could be adjusted for more mixed-ability groups.
AI Readiness Training — January 2026
Michelle Ko
Senior Analyst, Regulatory Consulting Firm
Working with Voxelith on regulatory filings extraction felt collaborative in a way that AI projects usually aren't. They wanted to understand why we needed certain data points, not just what. That curiosity made the resulting pipeline much better suited to our actual workflow. Delivery was on schedule and the handover was comprehensive.
Text Mining and Extraction — December 2025
James Wong
CTO, Professional Services Group
We needed an AI team that could work with our bilingual document environment — English and Cantonese mixed throughout. Most vendors we spoke to glossed over that complexity. Voxelith addressed it directly from the first conversation and built it into the pipeline architecture from the start. The result handles our mixed-language records well.
Knowledge Graph Construction — November 2025
Amy Lau
Learning & Development Lead, Corporate Bank
The training sessions were well-calibrated for a team with mixed technical backgrounds. Rather than assuming everyone was either a novice or an expert, the programme adapted to the room. The third session — on recognising where AI falls short — was particularly valuable for a team that had been using these tools with perhaps too much confidence.
AI Readiness Training — January 2026
Selected case studies
Challenge
Navigating a decade of fund documentation
A mid-size asset manager held ten years of research notes, meeting minutes, fund prospectuses, and investment theses across multiple formats and systems. Finding relevant precedent for a new analysis required hours of manual searching — and often produced incomplete results.
Solution
Voxelith built a knowledge graph connecting fund entities, companies, sectors, analysts, and time periods across the document corpus. The ontology was co-designed with the research team over two weeks to ensure it reflected how analysts actually think about relationships, not how a generic schema would model them.
Results
- Precedent search reduced from hours to minutes
- Cross-fund relationship queries now run in real time
- Onboarding new analysts accelerated by access to connected history
Duration: 11 weeks
Challenge
Manual contract review at scale
A legal operations team was reviewing hundreds of legacy contracts ahead of a regulatory compliance exercise. Each contract required manual identification of specific clause types, counterparty details, and key dates — a process that was both time-intensive and prone to inconsistency across reviewers.
Solution
Voxelith developed a custom text extraction pipeline targeting the twelve clause and data types the compliance team needed. A human review queue was integrated for cases where extraction confidence fell below an agreed threshold, keeping reviewers focused on the genuinely ambiguous cases rather than routine extraction.
Results
- Review time per contract reduced by approximately 70%
- Extraction accuracy on primary clause types: 91% precision
- Compliance exercise completed two weeks ahead of deadline
Duration: 5 weeks
Challenge
Building critical AI evaluation skills across a banking team
A corporate banking team had started using AI tools for drafting and summarisation but had no shared framework for evaluating the quality of outputs. Concerns were raised about inconsistent usage and an over-reliance on AI-generated content in client-facing materials.
Solution
Voxelith delivered the AI Readiness Training programme, adapted to the banking context with examples drawn from the team's actual use cases. Exercises used real AI outputs from their workflow, with the team practising structured evaluation and feedback techniques. The third session focused on designing internal review steps for AI-assisted materials.
Results
- Team adopted shared evaluation checklist post-training
- Confidence in evaluating AI outputs up significantly (self-reported)
- Internal review process established for AI-assisted client materials
Duration: 2 weeks (3 sessions)
Have questions before starting a conversation?
Ready to start a conversation?
A brief message is all it takes. We'll respond within one business day.
Get in Touch