Selling AI Is a Storytelling Problem, Not a Tech

There's a version of the AI pitch that gets made a thousand times a day in conference rooms across the country. It involves a slide deck, some benchmark numbers, maybe a live demo. It talks about efficiency gains and future-proofing. And it usually does not close the deal.

Nate B. Jones — an operator and AI strategist who has worked across Amazon and a series of venture- and private equity-backed startups — has a direct theory about why. He laid it out in a recent conversation with Nate Herk on the latter's YouTube channel: "If you are having trouble selling, the issue is that you are having trouble storytelling. You shouldn't be selling AI. AI is not valuable in and of itself. Sell that story. Sell the transformation. Sell the real impact of what you're delivering."

That's a deceptively simple diagnosis. The AI industry has spent considerable energy on capability demonstrations and considerably less on the question of whether the person across the table can actually picture the outcome in their own life. Jones argues this is the central failure mode — not the technology, but the narrative around it.

The Law Firm Problem

He offers a concrete illustration from the legal sector, where compliance concerns make AI adoption particularly fraught. A small law firm was evaluating multiple AI solution providers. Every vendor showed up with the same language: secure, compliant, certified. Table stakes, all of it.

One vendor won the business by doing something different. They brought a physical server into the room, set it on the table, and said: your data lives here and won't leave.

The technology behind that pitch may not have been superior to its competitors. The story was. Jones's point isn't that on-premise storage is the right answer for every law firm — it's that a story a busy, accountable decision-maker can instantly grasp will beat a features checklist every time. The server on the table was a prop in a narrative. The compliance slide decks were noise.

This applies with equal force to smaller operators. Jones uses the example of HVAC companies — businesses that struggle with phone coverage and dispatch logistics. The correct AI pitch to that business isn't about artificial intelligence. It's about answering the phone. "AI is not what you're selling," he says. "You're selling I pick up the phone."

The distinction sounds obvious until you watch how most AI agencies actually pitch.

The Mech Suit Mindset

Jones describes being "AI native" with a metaphor that's more useful than most of the jargon floating around this space. AI native builders, he says, think of themselves as wearing a mech suit at all times. The suit gives them capability across a wide range of problems. They reach for it first, second, and last — not because they've abandoned deterministic code or human judgment, but because the suit extends both.

What that looks like in practice: when facing an unfamiliar domain, an AI native operator doesn't stop and say they lack expertise. They map what AI-fluent practitioners in that space are doing, layer in what deep domain experts are doing with AI tools, and synthesize the two. Unsolved problems become intellectual property. The approach converts ignorance into a buildable artifact.

This is a meaningfully different frame than "use AI to work faster." It's closer to how one thinks about a power tool versus a hand tool — not just more efficient, but capable of things that weren't previously in reach.

The C-Suite Gap

The more uncomfortable argument Jones makes is about leadership. He agrees with the position that business leaders broadly need to become technically fluent — not just familiar with AI's existence, but actually using it. "When I sit down with CEOs," he says, "I tell them the buck stops with you. You are the AI change you want to see in the world. If you are not using the tools yourself, don't expect other people to."

This creates a real structural tension. The functions that most need AI transformation — marketing, HR, legal, finance — are typically the least technical parts of any organization. They've historically been served by IT departments and off-the-shelf software. The expectation that a chief people officer or chief marketing officer should develop genuine technical imagination about AI is a significant ask, and not one that maps cleanly onto most executive development pipelines.

Jones doesn't offer a tidy resolution here, and that's actually more honest than most of the commentary on this subject. The gap between what AI can do for a non-technical function and what non-technical leaders can envision it doing is real. Closing it requires something more than a licensing agreement.

The Token Distribution Problem

Inside companies that have already deployed AI tools, Jones describes a usage pattern that will ring true to anyone who has watched a corporate software rollout: a small cohort of passionate adopters consuming enormous amounts of AI capacity while a much larger group barely touches their accounts. He has observed this firsthand in teams of several hundred people — a handful of power users and a majority who are essentially paying for a tool they don't use.

His prescription is counterintuitive. Don't roll out more training. Don't issue more guidelines. Pick a challenge that is big, time-constrained, and slightly scary — something the team genuinely cannot deliver through normal means — and use it as a forcing function. That's when the organic knowledge transfer starts. People working shoulder-to-shoulder on something real start showing each other what's possible. The lights come on.

The other half of this picture is cost management. Jones is blunt about companies that respond to AI budget overruns by cutting access: "Did you not think about building an auto router? Send most of your queries to open models, hit the frontier when you need to, and watch your costs fall." The answer to AI cost problems, in his view, is almost never to ration the tool. It's to build smarter routing.

AI's Brand Problem and the Monster Stories

One of the more interesting threads in the conversation is AI's perception problem outside of tech circles. Jones observes — without much surprise — that the dinner-table sentiment about AI tends toward anxiety. He attributes part of this to the AI industry itself, which has a habit of amplifying alarming narratives about what the technology will eventually do while being vague about what it's actually doing right now.

He points to MidJourney — which, as Jones describes it, reached a substantial run rate with a small team, remained bootstrapped, and chose to reinvest in people and in medical imaging research — as the kind of specific, concrete story that counters the monster narrative more effectively than any reassurance does. He doesn't claim credit for verifying those details; it's his illustration of the kind of counter-narrative he thinks the industry should be building.

"We are telling ourselves monster stories about AI and we're telling them very loudly," Jones says. "And that's getting to our dinner tables. We need to be thoughtful and deliberate about the stories we tell."

That's a message directed at the AI industry as much as at any individual seller. The technology's reputation is partly a function of how its advocates talk about it.

The Hyperscaler Bet

Jones saves his spiciest position for last: he doesn't believe the future of AI will be determined by the major lab companies and their closed-source model benchmarks. He thinks it will be shaped by open-source models and by societies — he specifically mentions the UAE among others — that commit to AI adoption at a national level. His argument is that the "last mile" of implementation, the part that actually delivers value inside specific organizations and households, is something no hyperscaler can do for you. If they could, he notes, they wouldn't have spent so heavily trying to hire armies of deployment specialists.

Whether that prediction ages well is genuinely uncertain. Open-source model capability has been advancing faster than most people expected, but the resource requirements for training frontier models still favor entities with very large balance sheets. Both things can be true simultaneously: open models may handle most of the deployment work while closed labs continue to push capability boundaries. The question of who captures value in that arrangement is still open.

What's less debatable is the core observation Jones keeps returning to: the people who build real understanding of specific industries, specific workflows, and specific human problems — and who can translate AI's capabilities into those terms — are the ones with durable leverage. The model is not the moat. The story of what the model does for a particular human being, told to another human being, is the moat.

That's not a new insight about selling. It's a very old one, applied to a very new context. The question is whether the AI industry, which is not traditionally known for its narrative restraint, can actually execute on it.

Bob Reynolds is Senior Technology Correspondent at BuzzRAG.