Domain Knowledge & AI Discovery
One of the most underestimated constraints on enterprise AI adoption is not access to models, infrastructure, or data science skills. It is understanding the business well enough to know where intelligence could make a consequential difference. Most organisations do not suffer from a shortage of AI ideas. Ask a room of people for potential use cases and it will quickly produce a list containing customer service, fraud, forecasting, document processing, personalisation, and employee productivity. The harder question is whether those are the right opportunities for this organisation, in this part of its business, at this point in time.
Knowing the industry is not enough. Retail banking, for example, contains customer acquisition and onboarding, accounts and deposits, payments and cards, lending and credit, mortgages, customer engagement, and fraud and financial crime. Each domain contains different journeys, economics, decisions, data, and constraints. The same is true in healthcare, insurance, government, and other complex industries. Generic industry use-case catalogues can stimulate a conversation, but they cannot tell an organisation where its own performance is constrained or which decision is creating an unacceptable financial, customer, risk, or operational outcome.
I use a hierarchy of Industry → Segment → Domain → Journey → Decision to bring structure to this problem. The purpose is not taxonomy for its own sake. Each level adds business context until the discussion reaches a decision that can be examined in meaningful terms. A payment is not an AI use case. Within the payment journey are decisions involving authorisation, fraud, routing, liquidity, exceptions, and disputes. A claim is not an AI use case. It contains decisions around eligibility, coding, payment integrity, fraud, policy, and exception handling. Once the conversation reaches the decision, it becomes possible to understand what outcome is being influenced, what information is available, what is missing, and whether additional intelligence could materially improve the result.
Journeys are particularly useful because they connect activities that organisations often discuss separately. Consider a mortgage. Acquisition, application, underwriting, approval, servicing, delinquency, and refinancing may sit in different functions and systems, but they contribute to the economics of the same customer relationship. A conversation that begins with refinancing churn may therefore expose a more consequential opportunity elsewhere. Underwriting may be rejecting applicants who could have been profitable, early indicators of financial stress may exist before delinquency, or servicing interactions may contain information that could support earlier intervention. The initial use case becomes a point of entry rather than a boundary.
I have found this adjacent discovery particularly valuable. Clients usually arrive with a legitimate problem, but they naturally describe the problem they can already see. Understanding the surrounding journey makes it possible to ask whether another decision has greater financial consequence, greater frequency, or a more credible opportunity for improvement. It also makes it easier to determine whether AI is appropriate at all. Some poor outcomes are caused by missing intelligence, while others result from policy, process design, incentives, capacity, or organisational structure. AI should not be prescribed for a problem it cannot address.
Domain knowledge does not replace the client. The purpose is not to walk into an organisation and tell experienced business leaders how their business works. It is to arrive with enough context to ask better questions, understand the significance of the answers, and explore areas that might otherwise be missed. An experienced payments executive will always know more about the realities of their organisation than an external framework, just as an underwriting leader understands the policies, constraints, and exceptions that determine how decisions are actually made. Structured domain knowledge helps make that expertise easier to access and connect to potential AI opportunity.
There is also a scaling problem. No enterprise can place its best industry expert in every AI conversation, and no seller, consultant, or technical specialist can accumulate deep knowledge across every domain they encounter. If high-quality discovery depends entirely on finding the right combination of people for every meeting, AI adoption will remain inconsistent. Capturing domain knowledge through industries, segments, domains, journeys, and decisions provides a reusable structure that helps teams understand where they are in the business, what decisions typically occur there, which outcomes they influence, and what adjacent areas may deserve exploration.
As AI technology becomes easier to obtain, business understanding becomes more important, not less. The organisations that create the most value will not necessarily be those with the longest list of AI capabilities. They will be those that understand their own business well enough to know where those capabilities can change an outcome that matters.
