As enterprises across the region mark AI Appreciation Day on 16 July, a common thread runs through commentary submitted to us by senior technology leaders based in Singapore and across Asia Pacific: appreciation for AI’s progress is easy, but converting that progress into governed, production-grade value remains the harder task.

Five executives spanning cybersecurity, cloud, data management and enterprise software shared their views on where the region’s agentic AI ambitions are colliding with governance, data quality and operational readiness.

Cybercrime is moving at machine speed

ESET‘s Country Manager for Singapore and Asia, Pamela Ong, said AI Appreciation Day should also prompt a reckoning with how AI is reshaping the economics of cybercrime.

“Following ESET Research’s discovery of PromptLock, the first known AI-powered ransomware, JADEPUFFER shows how threat actors are using AI to automate more of the attack lifecycle. Human intent remains at the centre of cybercrime, but execution is increasingly happening at machine speed, giving defenders less time to detect and respond.”

Ong argued that AI security needs to extend beyond the model itself to cover the wider ecosystem — applications, cloud services, APIs, credentials and third-party integrations — with a prevention-first posture that catches abnormal behaviour before it becomes business disruption.

Agentic AI needs guardrails, not just power

At HPE, Fumiki Negishi, Vice President and General Manager for HPC & AI in the APJ GTM Division, said the region’s shift from centralised AI to edge inference is outpacing the governance needed to control it.

“The challenge now is not proving AI can work. It is scaling from isolated pilots to production-grade systems that can make real decisions on a real scale, without requiring human oversight at every step.”

Negishi pointed to Deloitte’s State of AI in the Enterprise report, which found just 21 per cent of organisations have a mature governance framework for autonomous agents even as adoption accelerates — a gap he called one of the defining challenges of the agentic AI era.

From AI ON the business to AI IN the business

Avanade‘s President for Asia Pacific, Bhavya Kapoor, drew a distinction between AI that merely speeds up existing work and AI that changes how a business actually operates.

“AI ON sits above existing work. It gives employees better tools for familiar tasks, but the same workflows, data issues and governance gaps remain. AI IN changes how the business works. It embeds AI and agents into workflows, decision models, governance structures and data foundations.”

Kapoor outlined five steps for enterprises to move from AI activity to AI value: starting with the process rather than the tool, making decision rights explicit before an agent goes live, embedding responsible AI into daily operations rather than policy documents, measuring business outcomes rather than usage metrics, and redesigning work around people rather than replacing them.

The data problem nobody wants to fix

Two separate commentaries converged on the same underlying issue: data readiness, not model capability, is what’s holding agentic AI back in production.

Informatica‘s Senior Vice President for Asia Pacific and Japan, Richard Scott, cited the company’s own CDO Insights research.

“Informatica’s CDO Insights research found that one in two organisations globally are using or planning to adopt agentic AI cite data quality and retrieval concerns as a major barrier to moving AI agents into production.”

Scott said the organisations that will get the most value from AI won’t necessarily be those spending the most on it, but those investing in the data foundations AI depends on.

Matthew Oostveen, Chief Technology Officer for Asia Pacific and Japan at Everpure, made a related point about the cost of getting this wrong.

“When AI is fed disconnected, context-poor data, it doesn’t produce intelligence – it creates dangerous hallucination risks and friction between human intent and machine execution.”

Oostveen said AI performs best when it can interpret data within its proper business context, and that scaling responsible AI means committing to what he called “the less glamorous work of getting our data in order.”

A common thread across the region

Taken together, the five commentaries point to a consistent gap in the region’s AI story: enthusiasm for agentic AI is outpacing the governance frameworks, data foundations and operating models needed to run it safely at scale.

For enterprises across Singapore and Southeast Asia, the leaders suggest, appreciating AI this year means being honest about that gap — and starting to close it.