Q&A
Which department should a company start AI adoption in?
Short answer
Do not pick a department because it sounds modern; pick it by criteria. Look for a process that is repetitive and high volume, data that already exists in usable form, and at least one person in that team genuinely eager to be the champion. Customer support, operations, or finance usually qualify first. Start small, measure, then expand.
The most common mistake I see: companies choose their AI pilot based on which department is the loudest or looks the most innovative. The result is a project that gets applause at kickoff and silence three months later, which makes management gun-shy about the next attempt. Choosing a pilot is not a popularity contest. There are three criteria, and they are simple.
First, a repetitive process with real volume. AI and automation deliver the most visible impact on work that repeats thousands of times. I experienced this directly when I drove customer support automation: from zero to about 70 percent of roughly 10,000 monthly tickets handled automatically, saving around USD 200,000 in 7 months. That happened before the generative AI era, and the math still holds today: large volume multiplied by a small improvement per transaction produces a big number.
Second, the data exists. Whatever process you want to put AI on needs a paper trail: ticket history, written SOPs, transaction logs, document templates. If the process lives entirely in one person's head and was never documented, AI has no material to work with. Clean that up first, then pilot.
Third, there is a champion. This is the most overlooked criterion and the most decisive one. You need at least one person in that department who actually wants this, not someone assigned against their will. The champion tests the AI output daily, feeds back what breaks, and becomes living proof for their colleagues. Without one, even the best tool becomes shelfware.
Apply those three filters and, in most companies, the answer lands on customer support, operations, or finance. Not because those departments are special, but because that is where all three criteria most often line up at once.
One closing note: start with one process, not a whole department. Define the metric before you begin, for example processing time or cost per transaction, measure for a few weeks, then decide whether to expand. A measurable pilot is what earns AI adoption the trust of management.
If your company is planning its AI adoption and wants a partner to design the pilot and upskill the team, take a look at my corporate program.