Key Insights from CAIS2026 Singapore: The Future of Conversational AI

Hosted by ConfX Global, the Conversational AI Innovation Summit 2026 in Singapore gathered senior leaders and tech professionals to outline the next phase of Conversational AI and Customer Experience. The consensus on stage was clear: the experimental phase of generative AI has ended. Today, enterprise leaders are moving past isolated pilots and focusing entirely on building governed, scalable AI infrastructures that deliver tangible business value.

Disclaimer: The views, strategies, and insights shared during the sessions reflect the independent expertise of the speakers and do not represent the official stances or policies of their respective organizations.

1. The Scaling Bottleneck: Governance and "Zero Trust"

A standout statistic from the event highlighted that 95% of AI pilot projects fail to reach enterprise scale. Naman Gupta, Head of Industry Solutions at SAS, pointed out that the main hurdle isn't the technology itself, but a lack of governance and trust. He recommended adopting a "Zero Trust" architecture, where organisations continuously verify data and restrict AI agents using least-privilege access.

Boomi Nathan, Regional Multicloud Architect at Prudential Services, built on this reality. In regulated industries, an AI system being "mostly right" is operationally equivalent to being "completely wrong". To build reliability and cut hallucinations by up to 80%, Nathan stressed the importance of Grounding (RAG). He also introduced the concept of "EvalOps" (Evaluation Operations), a process that uses an LLM as a judge alongside human-in-the-loop oversight to constantly validate AI outputs against a verified dataset before production.

2. Next-Gen Architectures: Graph RAG and Unified Intelligence

Traditional architectures are hitting their limits as enterprise use cases grow more complex. Terence Kok, Chief Innovation and AI Officer at Meinhardt, explained the transition from standard Vector RAG to Graph RAG. While standard RAG processes information in isolated chunks and struggles with complex reasoning, Graph RAG maps intricate data relationships to achieve over 97% accuracy. This makes it essential for strictly regulated fields like engineering and construction.

Akshay Seth from ComfortDelGro outlined another key shift: the need for a unified intelligence layer. Enterprises can enable autonomous actions by dismantling data silos and combining structured data, user behaviour patterns, and real-time conversation logs into a single framework.

3. Rethinking the Customer Experience: Generative UI

Vignesh Muthiah, Lead Solutions Architect at Singtel, offered a look into the next iteration of digital interfaces with Unified Generative UI. Currently, most chatbots rely on static text or rigid routing to display product options. Generative UI changes this by allowing the AI to dynamically build interface components, like specific forms or comparison tables, based entirely on a user's multi-part intent. To keep this process secure and maintain enterprise control, the AI operates within strict boundaries, requiring high confidence scores (e.g., 95%) before generating UI elements in real time.

4. Delivering ROI and "Return on Employee"

Sagar Rajput from FNO emphasised that organisations need to target specific use cases with clear ROI, noting that conversational AI can cut contact centre handling times by 30-35%. Harry Chan, Assistant VP of IHH Healthcare Singapore's IT Department, shared a practical case study from the private healthcare sector. His team launched two RAG-powered chatbots, "AskEinstein" and "AskAngelsX," to organise a chaotic internal document system. By synthesising thousands of clinical SOPs and administrative forms, these tools cut document retrieval time for nursing staff by 84%.

Taking a workforce-centric approach, Terence Kok championed the metric of "Return on Employee" (ROE). By using AI to shrink six weeks of heavy document review down to just a few days, companies can massively improve both productivity and overall employee satisfaction.

5. The Human Element: Co-Pilots, Not Replacements

Despite the push for automation, the human element is still critical. Eileen Ooi, President at PHD (a global communications and media agency), cautioned against "outsourced thinking," where teams blindly hand off complex evaluations to AI. She argued that AI should be used to challenge human perspectives, not just obey prompts.

During an expert panel moderated by Maria Singson, Founder & CEO at SMEMojo, Angel Lo, Director of Customer Experience (APAC) at Olympus, and Vinita Tandon, Executive Director at a leading BFSI, reinforced this dynamic. While AI handles lower-risk or repetitive tasks well, human specialists must remain the final decision-makers in high-stakes fields like medicine and finance. Finally, Jason Schadt, Head of Growth at Public AI, highlighted the growing "gig economy" of human experts who are compensated to annotate and train enterprise data, a process that is crucial for ethically sourcing AI models.

Building the Future, Together

If there is one defining takeaway from CAIS2026 Singapore, it is that technology alone isn't enough. Scaling conversational AI requires a unified ecosystem: robust governance frameworks, cross-functional architectures, and, most importantly, human expertise to guide and validate the models.

A Special Thank You to Our Partners

CAIS2026 Singapore was made possible through the generous support and collaboration of our network. We would like to extend our deepest gratitude to:

Our Exhibitor: Zapier and iZeno

Our Knowledge Partner: PublicAI

Our Community Partner: SMEMojo

Our Media Partners: AI Time Journal, AIPressRoom, Industry Events, Times of AI, DroomDroom, UToday, Unite.AI, Datafloq, and AI & ML Events.

Highlights from CAIS2026 Singapore

Want to see more from the summit? Check out some of our favourite moments from the stage and the networking floor below!