Golam Mustafa | Enterprise Technology Leader
1. You have spent nearly two decades working across IBM, SAP, Salesforce, AWS and enterprise transformation. What has this journey taught you about the difference between adopting technology and actually creating business value from it?
The biggest lesson is simple: implementing technology is not the same as transforming a business.
Over the years, I have worked with very different technologies and very different organisations. The technology keeps changing. The fundamental challenge does not.
A new platform can make a process faster. A new analytics platform can give people better visibility. AI can give us new capabilities. But none of that automatically creates business value.
The real question is always: what changed for the business?
Did we make a better decision? Did we reduce the time taken to act? Did we improve customer experience? Did we reduce working capital or operating cost? Did we help the business grow?
That is where I believe technology leadership has evolved.
Earlier, much of the conversation was about implementing systems. Today, it has to be about changing the way the business operates.
That is also why I say AI is not the strategy.
Technology is an enabler. The strategy comes from the business problem we are trying to solve and the outcome we want to create.
The real test of transformation is not whether the technology went live. It is whether the business became better because of it.
2. You have highlighted that “AI is not the strategy.” What should enterprises fix in their processes, data and decision-making before scaling AI initiatives?
Before bringing AI into the picture, I would first look at how the business actually works.
Many organisations have processes that have evolved over years. There are manual steps, multiple hand-offs, exceptions, duplicate data and workarounds that people have simply learned to live with.
AI does not automatically fix those problems.
If the underlying process is weak, AI can make a bad process faster. That is not transformation.
The same applies to data. Enterprises need reliable data, but they also need to understand where it comes from, who owns it and how it is used in a business decision.
Then comes the most important question: what decision are we trying to improve?
I would start there.
If we do not know which decision matters, what information supports it, who makes it and what happens when it is delayed or wrong, there is little point starting with an AI use case.
My approach is therefore straightforward: Fix the process. Strengthen the data. Understand the decision. Then decide where AI fits.
AI should follow the business problem, not define it.
3. As enterprises move from Generative AI toward Agentic AI, what do you believe will be the biggest shift in how businesses operate and make decisions?
The biggest shift will be from AI helping people with information to AI helping organisations take action.
Generative AI has already changed how people search, create, analyse and interact with information.
Agentic AI takes the next step. It can understand a situation, look at available information, identify an exception, evaluate possible actions and, within defined boundaries, execute one.
That has the potential to change how organisations operate.
Think about the current model. A problem occurs. Someone notices it. A report is generated. People discuss it. Someone decides what to do. An action is taken.
There is a lot of human effort between the signal and the action.
Agentic systems can potentially shorten that distance.
But I would be careful about giving machines unrestricted authority. Not every business decision has the same risk or consequence.
Some decisions can be automated. Some need approval. Some should always remain with a human.
So the real challenge is not simply building agents. It is deciding where autonomy makes sense and where judgement must remain with people.
The technology may become autonomous. Accountability cannot.
That will be one of the most important leadership questions as Agentic AI moves into the enterprise.
4. Many organizations successfully complete AI pilots but struggle to move into production. What are the biggest operational barriers preventing AI from scaling across the enterprise?
A successful pilot proves that something can work. It does not prove that an organisation is ready to operate it at scale.
That is where many AI initiatives struggle.
In my experience, the difficult part is rarely just the technology. The real challenge comes when the solution has to work with enterprise data, existing applications, security controls, users, processes and governance.
There is also a question of ownership.
If technology owns the AI solution but the business does not own the outcome, scaling becomes difficult. The business has to see the initiative as part of its operating model, not as another IT project.
Then there is the data problem. A pilot may work with a carefully prepared dataset. Production has to deal with the reality of enterprise data.
And finally, governance cannot be added at the end.
Security, accountability, access, monitoring and controls need to be considered from the beginning.
Before I scale an AI initiative, I would ask three basic questions: Does the business own the outcome? Is the underlying data reliable enough? Can we operate and govern it properly?
If those answers are not clear, the pilot is not ready for production.
The hard part is not proving AI can work. The hard part is making it work every day inside a real enterprise.
5. How should organizations identify the right business decisions and processes where AI or autonomous agents can create measurable impact, rather than simply adding AI to existing workflows?
I would not begin with an AI catalogue. I would begin with the business.
Look at the decisions that happen frequently, involve significant value, depend on large amounts of information, or suffer because people do not have the right information at the right time.
That could be demand planning, inventory, sales forecasting, customer prioritisation, pricing or exception management.
Then understand what happens today.
Who makes the decision? What information do they need? How long does it take? Where does the process get stuck? What is the cost of getting the decision wrong or making it too late?
That gives you a much better starting point.
The mistake is to take an existing workflow and simply insert AI into it because the technology is available.
Sometimes the right answer is not to automate the existing process. It is to redesign the decision itself.
That is where I see the real opportunity.
For me, the question is not: Where can we use AI?
It is: Where can a better decision create measurable business value?
Once that is clear, the technology choice becomes much easier.
6. What role do data quality, process maturity and governance play in determining whether an enterprise AI initiative succeeds or fails?
They are fundamental.
AI is only as useful as the environment around it.
If the data is unreliable, the output will be unreliable. If the process is poorly designed, automating it will not make it a good process. And if nobody knows who is accountable for a decision, adding AI creates another layer of uncertainty.
This is why I do not see AI readiness as simply a technology question. It is an enterprise discipline.
You need to understand the data. You need clear ownership. You need processes that are reasonably mature. And you need governance that is practical enough to protect the business without becoming a barrier to progress.
I have always believed that governance works best when it is designed into the way technology is delivered, rather than brought in after everything is built.
The same principle applies here.
Before giving an AI system greater autonomy, an organisation should know what data it is using, what decision it is influencing, what it is allowed to do and when it needs to escalate.
There is a simple way to look at it: Good foundations allow technology to amplify capability. Poor foundations allow it to amplify problems.
AI does not remove the need for discipline. It makes that discipline more important.
7. With AI increasingly influencing critical business decisions, how should leaders balance autonomy with human oversight, accountability and trust?
I do not believe the answer is to keep a human in every loop. That would defeat much of the value of automation.
The better approach is to understand the risk and consequence of each decision and then define the appropriate level of autonomy.
For low-risk, repeatable and reversible decisions, greater autonomy may make sense.
For decisions involving significant financial impact, customers, compliance or reputation, the level of human oversight should be higher.
The important thing is to define those boundaries before the system starts making decisions.
What can it decide? What can it execute? When does it need approval? When must it escalate? And who owns the outcome?
These are leadership questions, not just technical questions.
Trust will also come from transparency. People need to understand why a system reached a conclusion, what information influenced it and what controls exist around it.
I am comfortable with machines taking more decisions. I am not comfortable with organisations losing sight of who is accountable for those decisions.
Autonomy can increase. Accountability cannot disappear.
That is the balance leaders need to get right.
8. You have written extensively about moving from data movement to decision movement. How do you see AI changing enterprise decision-making from being reactive and reporting-driven to proactive and intelligent?
For years, enterprises have become very good at moving data.
We moved data from ERP systems into warehouses, from applications into data lakes, and from databases into dashboards.
But having more information does not necessarily mean making better decisions.
That is where I believe the next shift lies.
We need to move from data movement to decision movement.
Traditional reporting tells us what happened. Analytics helps us understand why. The next step is to identify what is likely to happen and what we should do about it.
That is where AI can make a meaningful difference.
Imagine a system that does not wait for someone to open a dashboard. It continuously looks at the business, identifies an exception, explains what may be causing it, estimates the likely impact and brings the recommended action to the person who needs to decide.
That is a very different operating model.
The goal is not to produce another dashboard. We already have enough dashboards.
The goal is to reduce the distance between a business signal, a decision and an action.
That is what I mean by decision movement.
And as systems become more capable, some of those actions can happen automatically within defined boundaries.
The future enterprise will not necessarily have more information. It will have a better ability to sense, decide and act.
9. From your experience leading digital transformation, what is one lesson you learned the hard way that today’s CIOs and technology leaders should keep in mind when driving AI-led transformation?
One lesson I learned the hard way is that a technology project can be successful on paper and still fail to change the business.
The system can go live. The integration can work. The reports can be available.
But if people continue working in the old way, if decisions do not improve, or if the business does not take ownership, the transformation has not really happened.
That changed the way I look at technology leadership.
I care about implementation, but I care more about what happens after implementation.
Are people actually using the capability? Has the process improved? Are decisions faster or better? Is the business seeing a measurable outcome? And who owns that outcome?
The same lesson applies to AI.
There is enormous pressure today to move quickly because the technology is moving quickly. But organisations should not confuse speed with progress.
The ability to build something quickly is not the same as the ability to create value from it.
My advice to technology leaders would be simple: Start with the business problem. Understand the decision. Build the foundation. Bring the business along. Then scale.
Transformation is complete only when the business works differently—and better.
10. Looking ahead, what will distinguish organizations that successfully become AI-driven enterprises from those that simply accumulate AI tools and experiments?
I think the distinction will be quite visible.
Some organisations will have many AI tools, pilots and experiments. Others will have changed the way they run the business.
The second group will win.
An AI-driven enterprise will not treat AI as a separate technology programme. It will become part of how the organisation makes decisions, serves customers, manages operations and responds to change.
But that does not mean using AI everywhere. It means being selective about where it creates real value.
The organisations that succeed will have clear business outcomes, reliable data, sensible processes, responsible governance and people who are willing to work differently.
Most importantly, their leadership will ask a different question.
Not: “What can AI do for us?”
But: “What should we now do differently because this capability exists?”
That is a much harder question. It forces leaders to rethink processes, decisions and sometimes the operating model itself.
I do not think the future belongs to the companies with the most AI. It belongs to the companies that know where intelligence matters, where decisions need to improve, and where technology can genuinely move the business forward.
That is the real test of becoming an AI-driven enterprise.
