Does this task need AI? A three-question test for clinics
Ask three questions in order. Does the information already exist in two places that are not linked? Connect them. Does the task follow the same rule every time? Automate it. Does it need a decision that changes case by case? Only then consider an AI agent.
Why start with the question and not the tool
Almost every organisation now uses AI somewhere, and few can point to money it made them. In McKinsey's 2026 global survey, 89% of respondents reported regular AI use in at least one business function, while 37% said it had contributed to profit. Only 6% reported a profit effect of 5% or more, and nearly three-quarters of that group had redesigned the workflow around the tool.
The pattern for small firms is similar. The OECD's 2026 survey of small and medium businesses in 12 countries found that 61% use at least one AI application and that 76% of those users are still at an early stage.
Neither survey covers Lebanon. They are still the best evidence available for a simple point: buying the tool is the easy part, and the return comes from fixing the workflow first.
Question 1: is the information already in two places?
A patient books on WhatsApp and someone retypes the booking into the schedule. The information exists; it is just not linked. The fix is a connection between the two, and it involves no AI at all.
Question 2: is it the same rule every time?
A reminder goes out 24 hours before every appointment. That is a rule, and a rule can run by itself. Many clinic systems already include this feature. If yours does, switch it on before paying anyone to build it.
Question 3: does it need a decision?
A patient cancels at nine in the morning. Who gets the eleven o'clock slot, and what do you say to the patient who cancelled? The answer changes each time. This is where an AI agent can help, as long as it has written limits and a person to hand over to.
Gartner predicted in June 2025 that over 40% of agentic AI projects will be cancelled by the end of 2027, citing rising costs, unclear business value and weak risk controls. Using an agent only where a decision is needed is how you stay out of that group.
What to do with the answer
Write down one task that bothers you and answer the three questions. If you stop at question 1 or 2, you probably do not need AI for it. If you reach question 3, check two more things before spending money: that the process is written down, and that the information it needs can be reached.