AI Oct 2026

According to TechCrunch, an Apollo economist argues that despite massive investment and widespread hype, AI has so far delivered "zero financial payoff" to corporate America outside the tech sector. That is one economist's assessment, not a final verdict on every company or every use case. Still, it puts a serious question to any manager: what exactly does our AI spending give back?

For an Iranian manager, this warning is not a reason to avoid AI but a reason to start with discipline. Technology budgets at many of our companies are tighter, and paying for foreign services usually comes with currency hurdles and access limits. In that environment, an unplanned purchase costs proportionally more than it would for a well-funded firm.

Step one is to drop the idea of an "organization-wide AI transformation" and pick one or two specific processes. Answering repetitive customer questions, producing periodic reports and analyzing sales are good candidates, because they recur and are easy to measure. A goal without clear boundaries will never show you its own results.

Step two is to define your metric before you begin. Measure the current situation for a few weeks: average customer response time, hours the team spends on the weekly report, or the error rate in sales forecasts. Without that baseline, you cannot tell whether anything improved or whether it merely feels that way.

Step three is a short, low-cost pilot, say a few weeks with a small team. Count the full cost, not just the monthly subscription: the rial equivalent of any foreign-currency spending, staff time for training and reviewing outputs, data preparation and security considerations. If customer data will be sent to an external service, decide in advance what information is allowed to leave.

Step four is to write down your decision criteria up front. Specify what result would lead you to expand the project, what would lead you to adjust it and what would lead you to stop. Ending a failed pilot is not a failure; it is a small cost that spares you a large mistake. If you do see non-financial benefits, such as better answer quality, record them with a defined metric too.

The takeaway is that the question isn't "AI, yes or no" but "start with numbers and evidence." One process, one metric and one short pilot are usually enough for an informed decision. If you would like a second opinion on choosing the process or designing the pilot, our team is glad to help.

Start a project