AI Won’t Fix Your Organisation: What Leaders Must Change First

Almost every conversation we have with business leaders in Malaysia eventually turns to AI, often starting with a discussion about which tools to use.

It is an understandable question, given the pressure leaders face to show their organisations are keeping up. But after years of working with companies to improve how they operate, we have come to believe that choosing the tool is rarely the right place to start.

The numbers help explain why. A 2026 study found that 24% of Malaysian workers are already among the most advanced users of AI, compared with 16% globally. That suggests employees in Malaysia are adopting AI quickly, but many organisations still need to catch up in how they redesign roles, processes and ways of working around the technology.

However, this pattern is not only unique to Malaysia. When MIT's NANDA initiative reviewed more than 300 enterprise AI projects, around 95% had yet to produce a measurable effect on profit. Without sound processes, AI risks automating confusion rather than solving problems.

For AI to deliver meaningful results, organisations must first examine how work gets done, where inefficiencies persist, and which decisions require human oversight.

1. Fix the process before trying to accelerate it

Many businesses are layering AI on top of processes that were struggling long before AI arrived. Approvals may sit across several chat groups. Decisions may depend on one experienced employee remembering what usually happens next.

AI does not make those problems disappear. In some cases, it simply allows a poor process to move faster. A customer request may take five days to approve because it passes through several people with no clear owner. AI can help draft a response in seconds, but it will not resolve the underlying delay.

The more important task is understanding why it takes five days in the first place. This means sitting down with the people doing the work and mapping out what actually happens, including which steps are simply carried over from the way things have always been done.

This is where the Japanese idea of kaizen becomes useful. Rather than looking for one big fix, it focuses on improving work through small, practical changes. We see the same principle when companies use Kintone to digitise everyday processes. Once teams can see where work is slowing down, they can remove unnecessary steps and keep refining the process over time.

2. Give people a problem to solve, and clear boundaries to solve it within

Another common mistake is telling employees to start using AI without explaining what they should be trying to improve. The same Microsoft study found that only 32% of Malaysian workers felt their leadership was clearly aligned on AI.

Without a clear objective, most will default to drafting emails or summarising documents, which saves time but rarely changes how the organisation operates.

Leaders can make experimentation much more useful by defining the business problem first. Instead of asking your finance team to use AI more, ask whether month-end closing can be reduced from eight days to three. Instead of measuring how many employees have adopted an AI tool, measure whether a process became faster, simpler or less error-prone.

Once the problem is clear, the people closest to the process are usually best placed to solve it. They know which approval always causes a delay and which temporary spreadsheet has quietly become essential to running the business.

At Kintone, we have spent years thinking about how much ownership people should have over the way they work. One lesson has stayed consistent: people are more likely to improve a process when they have genuine responsibility for it, rather than waiting for management or IT.

But autonomy still needs clear boundaries and accountability. AWS’s Malaysia study found that only 30% of businesses have clear accountability for AI, while just 18% have a documented process for what happens when something goes wrong.

Restricting experimentation may feel safer, but if the official process feels too slow, experimentation may simply move out of sight, including into public AI tools with information that should never have left the organisation.

A better approach is to define the boundaries from the start: what data AI is allowed to access, what employees are allowed to automate, who owns the result, and which decisions still require human approval. This is much easier when business information is already organised and permissions are clear.

This matters more as Malaysia’s National AI Office (NAIO) places accountability, safety and transparency at the centre of responsible AI adoption, making governance something to build into everyday workflows rather than treat as an afterthought.

3. SMEs may have an advantage, but only if they use it

Smaller businesses are often seen as having an AI advantage because they can move faster, and there is some truth in that. An SME owner might identify a problem on Monday, sit down with the people involved on Tuesday and begin testing a new process by the end of the week, while a large organisation may still be coordinating across several functions.

Malaysia’s National AI Action Plan 2026–2030 reflects this, identifying MSMEs as a priority and outlining efforts to reduce implementation costs and talent constraints.

But size alone does not make a business agile. SMEs can still be held back by outdated processes, disconnected spreadsheets and critical knowledge that sits with just one or two people. Their real advantage is the shorter distance between spotting a problem and changing the way work gets done. The businesses that make the most of it tend to start small: fixing one workflow, testing it with the people involved and improving it again.

This groundwork becomes even more important as AI moves from generating answers to carrying out tasks. AI agents can now update records, route requests and trigger the next stage of a workflow, yet a 2026 survey found that while 97% of executives had deployed AI agents in the previous year, only 23% were seeing significant returns.

The challenge is no longer adoption, but whether organisations have the processes and judgement to use AI well. Agents still depend on people to define how work should flow, where exceptions may arise and when a human needs to step in. That makes process thinking, judgement and accountability more valuable. Tasks can be delegated, but responsibility for the outcome cannot.

Start with the work, not the tool

So the next time the conversation turns to which AI tool to use, ask a different question first. Do we understand our own work well enough to improve it?

For leaders, AI readiness is less about having the latest technology and more about getting the basics right. That means improving the process first, giving people a clear problem to solve and setting the right boundaries around how AI is used.

As AI becomes more accessible, simply having the technology will matter less. The bigger difference will come from how well organisations use it to improve everyday work.

For Malaysian businesses, especially SMEs, that is an encouraging place to start. You do not need to transform everything at once. One better workflow, one clearer process and one problem solved well can already make a difference.

The organisations that get the most from AI will not necessarily be the ones using the most tools. They will be the ones that know where AI can genuinely help, and where better ways of working need to come first.