shares
Ask a room of business leaders whether they have tried AI, and most hands go up.
Ask how many have taken one of those pilots across the whole business, and the room gets quieter.
That gap sat underneath most of the discussion at the AI and Cybersecurity Forum on 27 August 2026, held at the Malaysia International Trade and Exhibition Centre in Kuala Lumpur. Organised by AIMX Malaysia ahead of the AI Malaysia World Expo and Conference 2027, the forum brought government, industry and academia into one room to look at where AI adoption actually stands.
The programme moved deliberately. National digital priorities in the morning, practical business application by midday, then cyber risk and resilience after lunch. That sequence says something on its own: adoption and security were treated as one conversation, not two.
Trying AI and Running on AI Are Different Projects
A pilot is built to succeed. It has a narrow scope, a willing team and a clean dataset. It answers one question: can this work?
Enterprise adoption answers a harder one. What changes when it does?
That second question involves people who never volunteered for the pilot, processes nobody has documented in years, and data spread across systems that do not speak to each other. None of it is a technology problem, which is precisely why a technology-led rollout stalls.
Organisations often read the stall as a failure of the tool. More often, it is a failure to define what the tool was supposed to change.
Why Pilots Stop at the Edge of the Business
The panel put the question directly. Many companies have started experimenting with AI, yet few have reached large-scale implementation. What is blocking them?
Three answers surfaced in different forms through the day.
Ownership
A pilot has a sponsor. Enterprise adoption needs an owner, someone accountable for the process itself rather than for the software running inside it.
Readiness of the Work
When a process is undocumented and handled differently by every team, automating it multiplies the inconsistency instead of removing it.
Workforce Capability
Adoption is a behaviour change. If the people expected to use the system had no hand in designing it, they will quietly route around it.
Governance Reads Like a Brake. It Works Better as a Route.
Responsible AI came up as a competitive question rather than a compliance one: how do organisations balance innovation with governance, data privacy and ethics while staying competitive?
In practice, the organisations that set boundaries early tend to move faster. Once the rules on data, decision rights and human review are written down, teams stop re-arguing them at the start of every project.
Skipping governance does not remove the delay. It moves it later, usually to the point where the system is already in production and the cost of changing course has multiplied.
The useful reframing for a leadership team is simple. Governance is not a tax on adoption. It is what makes adoption repeatable.
Security Belongs in the Design, Not the Cleanup
The afternoon turned to the cyber threat landscape in the AI era, and one observation applied directly to smaller organisations: many invest in digital transformation while treating cybersecurity as something to deal with afterwards.
AI adoption makes that ordering expensive. Every new system, integration and automated decision widens the surface that has to be defended. Adding controls after a tool is embedded in daily operations costs more than designing them in, and it usually arrives with an interruption to the business attached.
The practical version is unglamorous. Decide what data the system may touch before you connect it. Decide who reviews its output before it acts on anything. Handle both at design time, and cyber resilience stops competing with transformation for the same budget.
What Vision Brought to the Discussion
Vision joined two of the day’s sessions.
The first looked at AI transforming industries, exploring where generative AI, agentic AI and intelligent automation are producing measurable business value rather than promising it, and what those organisations did differently.
The second was the panel on AI in practice, covering implementation, governance, workforce readiness and the direction responsible adoption is taking in Malaysia.
Much of the value came from the questions the audience brought. Hearing where organisations are actually stuck is where any sensible starting point comes from.
What Vision Brought to the Discussion
For a leadership team deciding what to do next, the first step is not a platform decision.
Name the Business Problem
Not “we should use AI”, but “quotes take four days and we lose deals in the wait”.
Pick One Process
Preferably one that repeats, that people already dread, and where the result is visible within a quarter.
Design for the Operating Model
Decide who owns the process afterwards, what happens to the hours it frees, and how performance will be judged.
Set the Boundaries at the Start
Data, review, escalation. Agree on all three before the first system goes live.
Build Capability Alongside the System
Train the team as the process changes, so adoption does not rest on one person.
Then scale, because by that point there is something worth scaling.
Vision works with organisations at exactly this point: diagnosing where the business value sits, sequencing what to change first, and building the systems and the capability to make the change hold.
The Difference Is the Decision, Not the Tool!
The organisations pulling ahead are not the ones with the most pilots running.
They are the ones that decided, in business terms, what needed to change, and then used AI to change it.
Vision would like to thank AIMX Malaysia for the invitation, and everyone who joined the sessions for an open and practical discussion.
Explore AI for Your Organisation or reach us at hello@visiongroup.co
