Should Small Businesses Pay For AI Training Time
Thryv surveyed 561 U.S. small and midsize business owners and decision-makers. AI use reached 66%, but seven in ten said they need more training to use it productively. Most are cobbling that training together: 57% rely on YouTube and social media, 49% use online resources or webinars, and a third ask tools like ChatGPT how to use AI. That sounds cheap because the learning time is missing from the software price. In a five-person business, the office manager testing prompts after dinner, the marketer fixing a bad campaign draft, and the owner comparing tutorials are all paying for the rollout. Budget training before adding licenses: two paid hours on one real task, a short list of outputs nobody should trust without review, and one person the team can bring failures to. Then count those hours when deciding whether the tool saved money. If a $100 subscription needs ten hours of unpaid self-teaching every month, it is not a $100 tool. Should vendors include practical setup and training in the price, or should small businesses treat it like paid staff development?
Comments
Both, but not for the same part. The vendor should teach the buttons, data handling, and where the tool falls short. The business should pay for the local judgment: can a receptionist paste a customer’s name, which quotes need a second check, what never leaves the inbox. None of that should be homework on YouTube after closing. If the owner hasn’t set those rules, “training” just means each employee privately guessing where the risk is.
Put the training cost beside one finished task, not in a general AI-learning bucket. For four weeks, track baseline minutes, paid practice, review and correction time, and whether the customer reply, quote, or campaign actually went out right. The curve should bend: less checking and fewer fixes by week four. If it only works when the owner takes it home at night, the business hasn't adopted a $100 tool. It has hidden the rollout cost in the owner's evening.
Pay for the first failures too. If the receptionist catches a made-up refund policy, then spends lunch proving the tool was wrong, everyone learns to keep quiet next time. Give staff paid time to bring back one bad result and change the local rule. That is the part YouTube cannot do for the business.
The vendor also has to keep the lesson attached to the screen people actually see. A YouTube tutorial dies the moment a button moves or a cheaper plan hides a step. Give a new employee one harmless job in a practice copy of their own account: draft a customer reply, undo it, find the history. If they have to translate from somebody else’s dashboard, that isn’t training. It’s a scavenger hunt.
The press release tells us who answered, but not who absorbed the learning time. It surveyed 561 U.S. owners and decision-makers from a third-party panel, mostly at businesses reporting $1 million to $1.9 million in annual revenue. The 70% training gap and the expected 11–60 hours back are self-reports; the release does not measure employee training hours, wages, rework, or nights spent learning. Use that finding to justify a paid pilot, not to declare a return. Put one month of practice, checking, and fixes on the same timesheet as the hours supposedly saved. If the gain only appears after unpaid evenings, the business did not get time back. It borrowed it from somebody at home.