We do not understand machines very well because we are busy using them. Take a car, for example. We drive cars, ride in them and some of us sell them for a living. But what do I do if my car suddenly stops functioning? In most cases, you and I would call a mechanic. Even something as mundane as a tyre puncture needs a service provider. Ironically, we spend a lot of time reading about technology, it’s uses, issues and so on. But a trainer, manual, guide or even a YouTube video would not lead us to try and fix car issues on our own. When it comes to Artificial Intelligence, this is what I find. I am driving a new, shiny vehicle that keeps changing shape and capabilities and impresses me enormously. But if it malfunctions, I cannot fix it. If it refuses to deliver a service, I cannot do much about that. It may want access to my drive, and I can either consent or refuse. In some cases, the service provider of AI may offer a higher tier or different level of service that promises some more things, a better service experience, more computing resources and so on. But in all these cases, I do not understand the functioning of the machine. Nor can I go into its innards and do anything. With the privilege of time and especially if I am technically somewhat qualified, I might be able to understand a few underlying issues. But that is about it. There is little by way of agency for end-users when it comes to technology. When we talk about AI, that lack of agency has an impact. Why? Because AI is building both a baseline as well as a framework of intelligence, on its own. And this is where we come to the enterprise. A decision to use AI inside an enterprise is a different matter. Unlike the consumer, an enterprise owner needs to build, integrate and deploy services-and own the outcomes. The business is accountable to users, internal and external. In many cases, it may have potential liabilities, especially if it operates in a highly regulated industry. A consumer can simply decide not to use AI. Or not renew subscription. An enterprise owner sets himself or herself on a path of change with attached costs. This is why it is important for business owners to be AI literate. The starting point for this has to be data literacy. A recent project assessment I was involved in showed that a lot of data analytics does not need AI. Nor does it need heavy investment into infrastructure at the start. Yet there is a risk that decision makers may start discussing significant capex issues and application frameworks as solutions. In such cases, aggressive vendors may respond with a “I can do that”. That is appropriate and to be expected in business. But the impact on a business-and especially a relatively small one-may be disproportionate. One might be surprised (or not) but informal chats about what works and does not work are getting amplified in the case of AI. This is because of the speed of information dissemination and the pace of tech development. This calls for prudence and prudence needs knowledge empowerment. All business owners need to know that data pipelines are key to efficiency, internal governance transparency and the ability to see the future. But “data pipeline” is an abstract term for someone who might be spending a fourteen-hour day trying to keep a small firm running. So that burden of communicating needs easily, is one the technology community and especially the trainer sub-segment. Different industries and different functions have different data needs. A company may want to use AI to forecast its revenue takings for the next five years. Another may want to create a series of short movies for a vastly reduced cost. Both will have very different data pipelines. The very nature of the data and its attendant issues in each use case, will not be similar to the other use case, at all. Data pipelines and data analytics capabilities lead to the pursuit of automated predictability and automated functioning. This is possible without large language models, to an extent and depending upon the queries at hand. But once again, the dynamics of popular culture are such that conventional thinking starts with LLMs. It is easy to slide into as well, since there is a broad similarity (on the surface) to the way we consume apps and use search engines. The interplay helps of course. But use of prompts alone does not result in automation. There is significant integration work in coupling enterprise data, drives and computing stacks to LLMs. And there is considerable attendant risk. What is the balance an SME needs to strike? Are the costs and risks worth it? At what stage and business volume are these trade-offs worth considering? There are millions of small and medium enterprises (SMEs) in Southeast Asia and hundreds of thousands of mid-sized firms. In this part of the world, they collectively employ most people and contribute a large percentage to local GDP. If AI can provide efficiencies, strong predictive capability as well as creative capacity, then it will have enormous relevance. We know that AI Is embedded in much of what we do. It turns up in zoom calls, email summaries, searches and promos. We do not know how fast SMEs will adopt AI and integrate it at their enterprise levels. The extent to which this may occur is dependent to a large extent upon their owners understanding and being confident about it. It is not natural for vendors and service providers to ensure literacy about the building blocks of their industry. Much of that, due to the long innings of IT, is taken for granted. AI, however, is a different kettle of fish and a solid foundation of literacy and training is crucial. The time is now.
