Organisations that rush ahead to roll out agentic artificial intelligence (AI), without first ensuring they have the proper governance framework in place, likely will face roadblocks down the line.
Businesses feel the need to keep moving and not wait for agentic infrastructures to be ready before deploying, said Eileen Chua, SAP’s Singapore managing director.
This urgency has pushed some to start on AI projects, building agents on LLMs (large language models) on their own and often as a bolt-on to their legacy systems, said Chua, during a media briefing held on the sidelines of SAP Now AI conference in Singapore.
They do so because they cannot wait and need to demonstrate the value of AI to their business, she said.
Some of these enterprise customers now are realising that taking this route is not sustainable, without first looking at the governance framework they need to ensure their agentic workflows remain secure. Governance also is essential to facilitate key processes, such as access authorisation and authentication, she said.
Others that choose to start their AI journey early are pulling back in some areas where critical workflows are involved, in particular, financial transactions and sensitive customer data.
These organisations are taking the deliberate step to put in the necessary guardrails and have the frameworks in place before rolling out their AI agents, Chua said.
Agents have to be trusted and responsible by design, she noted.
This is essential especially for enterprise AI use cases, said Eric Wang, overseas digitisation head at TCL SunPower, at the media briefing.
While consumer AI use mostly revolves around creativity and mimicking a user’s thought process to produce content, businesses have to address issues related to accuracy and safety, Wang said.
Frontier models do a good job supporting consumer use cases, but are not yet adequate in facilitating enterprise AI, he said.
He pointed to smart manufacturing sites where a high degree of stability and control is critical to achieve accurate outcomes.
Agentic potential high, but ability to govern low
In Singapore, while organisations expect value from agentic AI, few feel they are ready to govern or scale the technology.
A majority 89% of Singapore organisations believe agentic AI has moderate to very high potential to transform their business, according to SAP’s Value of AI Report 2026, which polled 2,600 business leaders across 13 markets, including 200 in Singapore.
The expected ROI (returns of investment) from the technology will hit $9.8 million (SG$12.6 million) over the next two years, compared to last year’s estimated ROI of $4.9 million, the study found.
Despite the projected returns, however, only 2% of Singapore organisations are fully prepared for agentic AI, with 17% not prepared at all.
Some 56% believe they are partially prepared for agentic AI, while 26% say they are prepared with some gaps. the study revealed.
In addition, 66% of Singapore businesses already have piloted agentic use cases, but a lower 45% have a clear and shared understanding of what agentic AI is and what it can do.
Only 12% have the necessary skillsets, while 10% have the processes and frameworks ready to fully govern AI effectively, the study found.
Some 27% do not have a human-in-the-loop process for agentic workflows and 30% lack permission and access controls for agents.
Less than half, at 40%, have a registry of their agents.
In fact, 66% agree, or do not even know whether, they are deploying agents faster than they can govern them, according to the study.
Chua said: “Agentic AI raises the stakes for enterprise readiness. When AI systems can act across workflows, businesses need clear visibility into where agents are operating, what data they can access, and where human oversight is required. Without that foundation, organisations risk creating more activity without achieving better outcomes."
Get ready before jumping ahead
“Organisations pursuing AI strategies at enterprise scale face two challenges at once: risk that moves faster than most governance frameworks can keep up with, and value that is harder to measure than expected,” Sean Kask, SAP’s chief AI strategy officer, said in the report.
Chua added that value will not come from AI adoption alone.
“It will come from connecting AI to business data, processes, and governance, so organisations can move faster while still acting with control and confidence,” she said.
Respondents in the SAP study highlight four key areas with which they are struggling, she noted, pointing to the harmonising of processes, having data in the right places, governance, and talent.
She added that enterprise customers will not go live with new ERP systems if there are no AI use cases.
AI is a boardroom imperative, but companies need to flush out the areas that need more attention to be enterprise-wide ready, she said.
As AI becomes more capable, strong fundamentals become even more important, said Tan Kiat How, Singapore’s Senior Minister of State, Ministry of Digital Development and Information (MDDI), during his opening address at the SAP conference.
“They help us understand what AI is doing. Recognise its limitations. Challenge its assumptions. And know when human judgement matters most,” Tan said. “The foundations endure. What changes is what differentiates us.”
“Increasingly, the differentiator lies in how we combine these technical foundations with something more,” he said. “The ability to identify meaningful problems, understand customers, work across disciplines, combine technology with business and domain expertise, build systems that people trust, [and] translate technology into real-world impact.”
