Most small and medium-sized businesses in Malaysia understand AI will reshape their competitive landscape, but they're struggling to move from testing to real deployment. A 2025 study of 133 Malaysian SMEs found that just 21% had progressed beyond AI pilot projects, leaving 79% stuck in the experimentation phase. The Malaysian technology company Exabytes launched GROW AI at its summit in Kuala Lumpur to address this gap, aiming to support one million businesses in adopting AI by 2030.

What's Holding Small Businesses Back From AI Adoption?

The barriers to scaling AI are surprisingly consistent across Southeast Asia's SME sector. Around 60% of surveyed businesses cited a lack of in-house technical expertise as a major obstacle to moving beyond pilots. But technical skills alone don't explain the stall. Many organizations face a more fundamental problem: their data isn't ready for AI systems to work with.

"Many organizations want to implement AI and may even have the budget to do so, but one of their biggest challenges is that their data is not ready," said Chan Kee Siak, founder and CEO of Exabytes Group.

Chan Kee Siak, Founder and CEO, Exabytes Group

Data readiness means organizing and preparing information scattered across different systems, databases, and departments. Without this foundation, even well-funded AI projects struggle to deliver measurable results. This creates a catch-22: businesses need AI to compete, but they're uncertain where to begin and how to implement it securely.

How to Move AI From Individual Experiments to Company-Wide Impact

• Start with one operational problem: Rather than trying to transform every business function at once, identify a single, clearly defined challenge and build AI adoption from there. This approach reduces complexity and allows teams to learn before scaling.
• Prepare your data infrastructure: Before deploying AI tools, audit and organize data held across different systems. This foundational work is often overlooked but essential for AI systems to function effectively and deliver measurable financial returns.
• Measure at the company level, not the individual level: When AI adoption remains concentrated among individual employees, ROI becomes nearly impossible to track. Moving AI to company-level deployments with standardized metrics enables organizations to quantify business impact and justify continued investment.

Exabytes' GROW AI strategy is built on four principles: Reliable, Accessible, Scalable, and AI-driven, abbreviated as RASA. The framework reflects lessons from the company's own AI deployment. Exabytes' AI agent and avatar, named Eve, now handles more than 80% of the company's Level 1 workload, including sales support and product inquiries, after two years of refinement.

What Infrastructure Do Businesses Need to Scale AI?

Recognizing that different organizations have different computing needs, Exabytes announced several infrastructure options to support AI workloads at various scales. The company's high-performance computing and AI colocation facility in Penang is now operational, located within the region's semiconductor industrial area and designed for high-density computing workloads including GPU-based systems.

Beyond the Penang facility, Exabytes offers hosted Mac mini systems, DGX Spark instances, Exabytes Vision Cloud services, dedicated RTX PRO 6000-class GPU infrastructure, and managed Amazon Web Services (AWS) based AI services. These options span from smaller AI agent deployments and model inference to dedicated GPU computing and large-scale model training, allowing businesses to match their infrastructure investment to their current adoption stage.

Exabytes also signed a three-year partnership with Aurora Mobile, a Nasdaq-listed company operating the GPTBots.ai enterprise AI platform. The collaboration combines Exabytes' infrastructure and managed services with Aurora Mobile's enterprise AI technology to develop applications for Southeast Asian businesses. Initial use cases include customer service, sales, marketing, and knowledge assistants for SMEs, with larger deployments covering AI agents, service desks, workflow automation, and private AI environments for enterprises and public-sector organizations.

"Enterprise AI transformation requires more than deploying a standalone chatbot or application," said Chris Lo, founder and CEO of Aurora Mobile.

Chris Lo, Founder and CEO, Aurora Mobile

The partnership also emphasizes the importance of connecting AI technology with internal knowledge, workflows, employees, and operational systems while addressing governance and security requirements. This holistic approach reflects a broader shift in how enterprises think about AI: not as isolated tools, but as integrated systems that must align with existing business processes.

Eric Foo, deputy CEO of Exabytes, highlighted a critical challenge many organizations face when scaling AI: measurement. "The challenge is how to move the AI adoption from individual level, which is hard to measure the ROI, and elevate them to the company level with measurable financial metrics," Foo explained. This shift from grassroots AI experimentation to structured, measurable deployment is essential for businesses to justify continued investment and identify where AI delivers the most value.

The GROW AI summit brought together more than 500 business owners, executives, technology leaders, solution providers, and public-sector representatives to discuss practical AI implementation. Sessions covered cloud infrastructure, cybersecurity, workplace applications, commerce, workforce development, customer engagement, data protection, human resources, and organizational effectiveness. This breadth reflects the reality that AI adoption isn't a single technical problem; it's an organizational transformation that touches nearly every business function.

As Malaysia pursues its AI Nation 2030 agenda, initiatives like GROW AI signal a shift from experimental AI projects toward systematic, scalable deployment. For the 79% of Malaysian SMEs still in pilot mode, the path forward requires addressing data readiness, building technical expertise, and moving from individual AI use to company-wide measurement and governance. The infrastructure and partnerships now available suggest that the barrier to scaling AI is no longer primarily technical,it's organizational.