How to Start Adopting AI When You Don’t Know Where to Begin

Practical AI#3

Most organizations don’t have a technology problem. They have a clarity problem.

One of the most common questions I get from executives isn’t about which AI vendor to choose or how much to budget. It’s more fundamental than that:

“We know we need to do something with AI. We just don’t know where to start.”

What makes this question harder than it sounds is that there often isn’t a ready-made answer to copy. In many markets — particularly across Southeast Asia — genuine, detailed case studies of enterprise AI adoption are still scarce. Even the major global consulting firms, when you push past the polished frameworks, are often offering repackaged common sense rather than hard-won operational insight.

So here’s the starting framework I give executives when they’re standing at the beginning with no map: Know the technology. Know your organization. Then bring them together.

Know the Technology — Before You Trust It

Most executives I work with have seen impressive AI demos. A few have been genuinely wowed. Some have been burned.

The problem with demos is that they’re optimized for the best-case scenario — clean inputs, predictable queries, ideal conditions. Real organizational data is messier. Real use cases are more ambiguous. And real AI systems, even the best ones available today, hallucinate, misinterpret context, and fail in ways that don’t show up in a vendor presentation.

The fastest way to close this gap is to simply use the tools yourself.

Not in a structured evaluation. Not through a procurement committee. Just open ChatGPT, Gemini, or Claude and start pushing on them. Try to break them. Feed them the kind of documents and questions your organization actually deals with. Notice where they’re surprisingly capable and where they fall apart unexpectedly. Learn from a lot of readings that do not only discuss successful use cases, absorb information about limitations and some failed cases too.

This matters because executives who’ve never personally used these tools tend to miscalibrate in both directions — overestimating AI’s reliability in high-stakes decisions, and underestimating how much operational lift it can provide in repetitive, well-defined tasks.

Firsthand experience won’t make you an AI expert. But it will make you a much better buyer, a much sharper client, and a much more effective decision-maker when it comes time to commit resources.

Know Your Organization — Before You Commit to a Direction

Once you have a realistic model of what AI can and can’t do, the second step is turning that lens inward.

This means doing an honest audit of your own operations — not from a strategy document, but from the ground up. A few questions worth sitting with:

Where is time actually going? Not where you think it’s going based on org charts and job titles, but where hours genuinely accumulate on the ground. The answer often surprises leadership.

Where do errors cluster? Manual, repetitive, high-volume processes tend to be both the most error-prone and the most amenable to AI-assisted improvement. That overlap is usually where the early ROI lives.

Is your data actually ready? AI needs data at the right time, in the right format, with the right access controls. Many organizations discover, only after committing to an AI project, that their data infrastructure isn’t set up to support it. This is worth assessing honestly before you’re mid-implementation.

What’s the change appetite of your people? Technology is rarely the binding constraint in AI adoption. Culture and incentive structures usually are. An organization whose frontline staff feel threatened by automation will resist AI in ways that no vendor roadmap accounts for.

Why Copying Others Usually Falls Short

There’s an understandable temptation to look sideways — to find out what a competitor or an industry peer is doing with AI and replicate it.

The problem is that organizational context isn’t portable.

What works for a large retail bank with thousands of standardized loan applications may not translate to a mid-sized life insurer with complex underwriting decisions and long policy cycles. The surface-level use case might look similar. The underlying process, data structure, regulatory environment, and workforce dynamics often aren’t.

Benchmarking has its place. But it’s a starting point for questions, not a substitute for the harder work of understanding your own situation.

The Underlying Logic

Know the technology well enough to have realistic expectations. Know your organization well enough to identify where those capabilities genuinely meet your needs.

That intersection — between what AI can actually do and what your organization actually needs — is where sustainable AI adoption begins.

Everything else, the vendor selection, the implementation approach, the change management — flows more naturally once that foundation is in place.

Part of the Practical AI for Business series — practitioner perspectives on enterprise AI adoption.

#PracticalAI #AIforBusiness