Government Embraces AI: Sarawak Accelerates Voice AI Rollout for Public Services!

Sarawak Information Systems Sdn Bhd (SAINS) and the Malaysian AI company 8nabler are embarking on additional voice AI trials, following a successful two-month proof of concept (PoC), to determine the feasibility of a commercial deployment across government services. The initiative stems from a Memorandum of Understanding (MoU) signed at the Digital Government Exchange 2026 in Kuching in September, which established a framework for exploring voice AI applications within SAINS-supported government systems and public-facing services.

The initial PoC involved SAINS’ internal ICT helpdesk, during which the AI system processed 4,056 calls, with 1,225 identified as genuine civil-servant ICT incidents. According to 8nabler CEO Benjamin Foong, the system adeptly handled requests in mixed Bahasa Malaysia and English, classified reported problems, and routed them to specialized queues covering hardware, software, and network services. This trial, however, was limited to SAINS staff and existing telephone infrastructure, without extending to other government agencies or public-facing services.

Despite its successes, the initial testing phase revealed technical limitations, particularly concerning the recognition of regional accents, government-specific terminology, departmental names, internal applications, and acronyms not adequately represented in its original training data. 8nabler’s partnership documentation specifically highlights local vocabulary and scripted interactions as areas requiring further refinement for optimal performance.

The upcoming phase of trials aims to address these identified limitations and assess more complex applications. While a precise timeline for conclusion and the specific government systems to be evaluated were not disclosed, Foong confirmed that objectives and a timeline have been established. A key focus will be on expanding language support to include Iban, Mandarin, and Tamil, alongside mixed-language conversations, which is crucial for a diverse region like Sarawak.

A significant component of the next phase involves evaluating data governance and human oversight. This includes assessing whether AI models, voice processing systems, and conversation records can be hosted within Sarawak, along with implementing robust data access, retention, and deletion controls managed by government agencies. Proposed safeguards include masking personally identifiable information (PII) before it reaches AI models, maintaining comprehensive audit trails of conversations, and restricting data access based on agency roles. These considerations are in line with Malaysia’s National Digital Department (JDN) Public Sector AI Adoption Guidelines, which cover responsible AI use, data privacy, security, transparency, accountability, and system reliability.

Regarding human oversight, 8nabler’s Panacea platform utilizes confidence thresholds, predefined rules, and approved information sources to decide when to respond directly and when to escalate a request to a human agent. Escalation occurs if the AI cannot provide a sufficiently confident answer, repeated responses fail to resolve an issue, or the request involves sensitive information or falls outside its authorized scope. Foong identified information requests, status enquiries, and appointment bookings as suitable initial public-facing applications, emphasizing that decisions involving welfare eligibility or individual rights will continue to necessitate human judgment.

Beyond internal ICT assistance, SAINS and 8nabler are also exploring wider public service deployment. This involves evaluating integration with existing government applications and assisted-service channels. A proposed use case would allow residents to access information or assistance through ordinary telephone calls, bypassing the need for mobile applications or online forms. This backend AI processing would enable access for users without smartphones or personal broadband connections, though telephone network coverage would remain a prerequisite. Future evaluations will also examine the technology’s ability to connect with applications from different government agencies and handle increased call volumes during peak demand.

Furthermore, the partnership will investigate converting conversation records into structured data, which could empower agencies to monitor service demand and identify gaps in service delivery. Despite the advanced planning and ongoing trials, SAINS and 8nabler have not yet announced a commercial agreement, a public rollout date, or the specific government services that will be included in the next phase of testing.