A common first conversation about AI begins with a product name and a question: should we be using this? It is a reasonable question with no useful answer until the business says what it wants to change. A tool is a solution, and a solution needs a problem to be measured against.
Starting from the tool also narrows the thinking. The team looks for places to use what it has seen, instead of asking where the business loses the most time and money.
List the work, not the wishes
Ask each team to write down the tasks that take the most time each week and that they would happily never do again. Be specific: reconciling supplier invoices, answering the same ten questions from dealers, drafting product listings in three languages.
A list like this is more valuable than any strategy slide. It shows where hours go, and it is written by the people who will judge whether anything improved.
Pick one workflow and draw it
Choose a single task from the list and map it step by step, from where the input arrives to where the output goes. Note who touches it, what they check and where mistakes happen today.
Often the map alone reveals that part of the problem is not an AI problem at all. A missing form field or an unclear handover can cause more delay than the task itself.
Decide what better means first
Before trying anything, write down how the task is measured now: time taken, errors caught later, how often it waits in someone's inbox. Agree what a worthwhile improvement would look like. Without a baseline, every pilot looks promising in a demo and nobody can say whether it paid for itself.
Run a small trial against the old way
Run the new approach on a sample of real work for a few weeks, alongside the existing process, with the same people checking the results. Compare the numbers you wrote down, and ask the team how it felt to use.
If it clearly helps, extend it. If it does not, you have spent a few weeks, not a year, and you know more about the workflow than you did.
Keep the people doing the work in the room
The staff who do the task every day know its exceptions: the supplier who always sends scanned invoices, the dealer who writes in Kannada. Involve them in the trial and in judging it. A tool they helped shape is a tool they will actually use.
