Thawn AI
Client feedback

What our clients
actually say

We keep a record of client feedback from every engagement. The views shared here reflect the range of projects we've completed — including ones where things didn't go perfectly.

Back to Home
40+

Completed engagements

4.8

Average satisfaction score (out of 5)

87%

Clients who returned for a second engagement

4 yr

Years delivering AI services in Malaysia

Reviews

Client testimonials

Selected feedback from clients across different service types and sectors.

NK

Nurul Kamaruddin

Operations Manager, Kuala Lumpur

We came to Thawn with very little clarity on what we actually wanted from AI. The scoping service gave us something concrete to bring to our board for approval. The brief they produced was thorough and written in language our non-technical stakeholders could understand.

AI Project Scoping · March 2026

TW

Tan Wei Liang

Head of IT, Selangor

The email triage system has been running for two months now. The time our customer support team spent manually sorting mail has dropped noticeably. The integration with our existing platform was smoother than expected, and the documentation made maintenance straightforward.

Email Triage · February 2026

SA

Salmah Ahmad

Product Lead, Kuala Lumpur

Honest feedback: the multi-modal project ran about two weeks over the original estimate due to some complexity we hadn't anticipated with our data setup. Thawn flagged it early and worked through it transparently. The final result is solid and our team now has a clear picture of how it works.

Multi-Modal Integration · January 2026

RH

Razif Harun

Managing Director, Penang

We'd spoken to a few larger vendors before Thawn, and they all came back with enormous proposals. Thawn's scoping engagement helped us understand that we actually needed something much more targeted. The brief saved us from committing to a project that would have been far too large for where we are.

AI Project Scoping · March 2026

CY

Chan Yu Ming

Senior Analyst, Klang Valley

Communication throughout the email triage project was consistent. We had a scheduled check-in every week and there were no surprises. The team was direct when they needed input from our side and clear about what they needed and why.

Email Triage · February 2026

FL

Faridah Lim

COO, Johor Bahru

The multi-modal system handles our document intake and image classification in a way we couldn't have built in-house. It's not flashy but it does what it's supposed to. The handover documentation is genuinely useful — we've referred back to it several times since.

Multi-Modal Integration · January 2026

Case studies

Selected engagement stories

A closer look at three engagements — what the client needed, what we built, and what happened.

Professional services firm — Email Triage

Kuala Lumpur · 6 weeks · Completed March 2026

The challenge

A Kuala Lumpur-based firm receiving 400–600 inbound emails per day across multiple practice groups. Staff were spending 1–2 hours each morning triaging before substantive work could begin.

The solution

Classification system trained on 6 months of labelled historical mail. Integrated with their Microsoft 365 environment. Handles sorting, urgency flagging, and routing to practice-group inboxes.

The outcome

Manual triage time reduced from approximately 90 minutes to 20 minutes per morning. The system now handles 78% of incoming mail without human review. Classification accuracy sits at 91% on unseen data.

"The time saving was real and immediate. More importantly, our staff now start the day doing the work they were hired for."

— Operations Manager

Property tech startup — Multi-Modal Integration

Petaling Jaya · 12 weeks · Completed February 2026

The challenge

A property platform needed to analyse listings from multiple sources — including text descriptions, floor plan images, and structured pricing data — through a single interface for their recommendation engine.

The solution

Multi-modal pipeline combining a text encoder, a vision model for floor plan analysis, and a tabular data processor. Unified API with a single endpoint for the client's existing backend to call.

The outcome

Recommendation relevance score (internal metric) improved by 23% compared to their previous text-only approach. API response time averaged 340ms across test cases. Now live in their staging environment.

Manufacturing SME — AI Project Scoping

Shah Alam · 2 weeks · Completed March 2026

The challenge

A Shah Alam manufacturer had been approached by several vendors proposing AI-driven quality inspection systems. They had no internal expertise to evaluate the proposals and wanted clarity before any decision.

The solution

Scoping engagement covering their production line data, defect patterns, camera infrastructure, and team technical capacity. Included an independent assessment of two vendor proposals they had received.

The outcome

Client received a detailed project brief and a recommendation to proceed with a targeted pilot — not the full deployment either vendor had proposed. Decision-making time reduced significantly with a clear written basis for board review.

Certifications and affiliations

MDEC Digital Economy Programme

Participant, 2023

Malaysia AI Community

Active contributor

NASAM Corporate Partner

Pro bono AI scoping, 2024–2025

Reach the team

Your project

Ready to see what's possible for your team?

Get in touch for a straightforward conversation about your situation — no commitment required.