Initially, most attendees of a Fortune Brainstorm Tech conference panel in Aspen, Colo. on Tuesday agreed with the assertion that all companies need an AI game plan. But that consensus withered after some discussion.

For small companies, in particular, it probably makes no sense to dedicate limited resources to hire an AI expert, even if there were one available. It was also unclear how a mom-and-pop business, say a tailor shop, could benefit from AI, although Zachary Bogue, co-managing partner of the Data Collective, an early-stage tech venture fund, said that robots can already handle and sew fabric.

“Maybe every industry needs an AI strategy, but not every business,” said George Kurtz, CEO of CrowdStrike, a cybersecurity specialist.

Related: Beware the Hype of AI

“AI is just beginning, so having a strategy around it is a problem because you have to define what you’re talking about,” said Norman Winarski, founder of Winarski Ventures. “A lot of industries will not do well by deploying chatbots,” he added, referring to those automated customer service agents that are increasingly used by companies online. “You have to be incredibly careful in how you deploy an AI solution, you need to think about how people will react, and it takes a lot of resources.”

And small businesses, as noted, have limited resources. And, given that small companies are the vast majority of American businesses, this is a rather large gap in the AI picture.

Get Data Sheet, Fortune’s technology newsletter.

Having said that, it’s likely that many small businesses will end up using software that incorporates AI, said Hilary Koplow-McAdams, former president of New Relic (NEWR) and now board member of Tableau (DATA). Virtually every business software company from Salesforce (CRM) to Microsoft (MSFT) says they are weaving AI capabilities into their products.

Several attendees were skeptical of the notion that all “smart” applications fall under the AI umbrella. Expert systems technology that emulates human decision processes doesn’t necessarily require a lot of data, for example, said Michael Schrage, a fellow at MIT’s Center for Digital Business.

While machine learning, by its nature, requires large amounts of data in order to learn to recognize patterns, there are other jobs in which simple rules or operations can be automated without a lot of data.

“If you compete just on massive data, you’ve already lost to Facebook (FB), Google (GOOG), and Microsoft (MSFT) you’ve already lost,” he said.

And, in those cases where a company does have a lot of data at its disposal, that data isn’t all that valuable for AI if it is not formatted and prepared correctly. Several attendees stressed the need for “data refineries” to clean that data.

“If your data is not cleansed, you cannot use AI,” Kurtz said.

So in this new-age era of AI, the old maxim still applies: Garbage in, garbage out.