Is your AI saving hours or actually making you money?
Most businesses use AI, but few can point to the profit. A simple cost-versus-earnings check for each AI task can tell you which tools are paying off.
What happened
Most organizations now use AI, but far fewer can show it is making them money. In a piece for Forbes, Vaibhav Dani, CEO of Map Communications, cites McKinsey's 2026 State of AI survey. It found that nearly nine in 10 organizations use AI regularly, yet only 37% credit it with any impact on their earnings.
Dani also stresses that AI has a running cost. According to Forbes, every prompt, multi-step workflow, reasoning cycle and lookup request uses computing power, and the article calls tokens a new variable cost for companies. In plain English (our explanation, not the article's), tokens are the small chunks of text many AI tools count when they bill you. Traditional software cost almost nothing extra each time someone used it, the article says. AI does not work that way.
Why it matters
Saved hours sound great, but they are not money yet. Forbes carries Dani's example: if an assistant saves each employee five hours a week, the value only appears when that time produces revenue, improves service or avoids a hire. Otherwise, he writes, the savings may never reach the income statement.
Falling prices don't solve this either. According to Forbes, Andreessen Horowitz has documented the cost of running equivalent AI models falling roughly 10 times per year, yet consumption is rising faster. So the total bill can still grow even as each unit gets cheaper. Dani's answer is to judge AI by the return on what it costs, not by how many people have logged in.
What it means for your business
- Restaurants and retail shops: If AI answers customer messages or writes product descriptions, count what each task costs and whether it brought in orders, repeat customers or fewer missed inquiries.
- Law firms: If AI summarizes documents, ask where the freed time went. Billable work, faster turnaround and avoided overtime count. Lighter afternoons do not.
- Agencies: Track whether AI lets you take on more clients without adding staff at the same pace. That is one way to see if the tool is paying for itself.
- Clinics: If AI handles scheduling questions or reminders, measure cost per resolved question and compare it with the staff time it replaced.
In the same Forbes article, Dani suggests measuring automation by cost per transaction, and sales or personalization tools by extra revenue and conversion. The article is written for big-company CEOs, but you can try the same idea on a smaller scale with a simple spreadsheet.
One thing to try this week
Pick one AI task you use regularly and write down three things on a single sheet of paper:
- What it costs per month (subscription plus any usage fees).
- How many hours it frees up.
- What those hours were spent on.
If you can name a result, such as more booked jobs, faster invoicing or a hire you did not need, keep the tool. If the answer is "not sure," you have found the one to review first.