AI in Marketing: The Three Buckets

Everyone is asking how to use AI in marketing. It is the wrong question, because there are three completely different answers, and teams that blur them end up chasing all three badly instead of winning the one that matters for their business.

I run all three buckets today, hands-on, in businesses I own or have operated. Here is how I separate them, and how to decide where your effort should go.

Bucket one: internal tools

This is the AI your team uses to work faster. The LLMs, the scripts, the agents that draft, analyze, summarize, and build. It is the bucket everyone means by default when they say “we are using AI,” and it is the easiest to start and the hardest to do well.

The trap is tool collecting. A team with eleven AI subscriptions and no changed workflows has adopted nothing. The test I use is simple: name a process that runs materially differently today than it did two quarters ago because of an AI tool. Not faster first drafts. A changed process, with a step removed, a handoff eliminated, or an output that did not exist before.

At Upgrade Labs, my human performance business in Park City, the reporting command center, the financial forecasts, and the commission calculations are built and maintained by AI agents I direct. That is what bucket one looks like when it lands: not a subscription, a system. The work still needs an operator who knows what good output looks like, because agents amplify judgment rather than replace it. But the lift is real, and it is the closest thing to free capacity a small team can get.

Bucket two: customer-facing AI

This is the AI you build into the product and experience your customers actually touch. Recommendations, assistants, personalization, AI features in the product itself.

This bucket carries the highest stakes, because your customers experience the failures directly. An internal tool that hallucinates wastes an afternoon. A customer-facing feature that hallucinates damages trust you spent years building. The bar for shipping here should be product-grade, not pilot-grade.

It is also the bucket where the strategic payoff lives. Internal tools make you cheaper to run, but your competitors have access to the same tools, so the advantage erodes. Customer-facing AI, done well, changes what your product is. The question to ask before building anything here: does this feature make the customer’s outcome better, or does it make our company look current? Only one of those survives contact with a renewal decision.

Bucket three: platform AI

This is the AI your existing stack is quietly shipping. Your CRM, your marketing automation platform, your ad platforms. Most teams are paying for capabilities they never turn on.

This is the most neglected bucket and often the highest immediate return, because the integration work is already done and the cost is already in your budget. The ad platforms have been the clearest case in my world: campaign types and optimization features that materially change performance, sitting behind a settings page nobody has audited since the contract was signed.

The move here is an audit, not a purchase. Once or twice a year, have someone inventory what your existing vendors have shipped and decide deliberately what to enable, test, and ignore. An afternoon of reading release notes routinely beats a quarter of new-tool evaluation.

Pick your wave on purpose

The teams that win with AI are not the ones chasing every new tool that trends on a Tuesday. They are the ones who pick the wave that matters for their business and ignore the rest deliberately.

A ten-person DTC brand probably lives in buckets one and three: agent horsepower for the team, and every capability the ad platforms will give them. A funded product company with engineering depth should be fighting in bucket two, because that is where moats form. A services business might find that bucket three alone changes its economics.

The noise is relentless right now, and the noise does not care about your P&L. Sort every AI decision into its bucket, fund the bucket that compounds for your business, and let the other two run on maintenance effort. Intentional beats current. It always has.

For what this looks like in production, read the owner-operator case study, or get in touch.