Almost everything in life has a tradeoff. Fire cooks food and burns down villages. The automobile connected the world and reshaped it around collisions no one had a word for yet. Every meaningful technology arrives carrying both of these at once — and the gap between when the advance shows up and when society figures out how to hold it responsibly is usually where the damage happens.
I’ve watched this pattern up close once already. I’d like to explain why I’m not willing to watch it happen the same way twice.
I saw the upside of social media. I missed the cost.
When social media emerged, I understood immediately what it meant for marketing. I could see how a skilled marketer with a real strategy could use these platforms to reach a massive audience and move it — build product adoption, build brand favorability, build loyalty at a scale that hadn’t existed before. I was right about that, and it’s mostly what I paid attention to.
What I was naive about was everything else it would do on the way there.
We now have a substantial body of evidence that heavy social media use is associated with real harm to mental health, particularly among adolescents. The U.S. Surgeon General’s 2023 advisory found that children and adolescents who spend more than three hours a day on social media face roughly double the risk of experiencing symptoms of depression and anxiety — and that up to 95% of teens report using a social media platform at all, with a third saying they use it “almost constantly.” Pew Research has since tracked the shift in how teens themselves see it: in 2022, 32% of teens said social media has a mostly negative effect on people their age. By 2025, that number had climbed to nearly half. The people living inside the platforms are telling us, in growing numbers, that the platforms are hurting them.
I don’t think anyone building those platforms sat down and designed for that outcome. I think it’s what happens when an enormously powerful technology gets deployed at ubiquitous scale before anyone has done the work of guiding it into how people actually live — our norms, our culture, our psychology, our behavior.
That is the mistake I don’t want to make again.
AI is the same pattern, except faster and far more powerful
Technology innovation isn’t something to resist. It’s something to accept — and then do the harder work of shaping so it strengthens rather than erodes the norms, culture, and behavior we already have. Nothing this powerful can be handed to everyone, everywhere, with no guidance on how to use it, and be expected to land well by accident. That’s true of social media. It’s true of AI, at a different order of magnitude.
AI is genuinely extraordinary — for helping and for hurting, often through the exact same mechanism. The difference between those two outcomes is almost never the model. It’s whether a human being was actually driving.
We’ve already watched this play out inside real companies. Microsoft’s Tay chatbot was manipulated into offensive output within a single day of unsupervised public interaction. Amazon quietly killed an internal recruiting tool after discovering it had taught itself a bias against women from a decade of historical resumes nobody had checked closely enough. Zillow shut down its algorithmic home-buying arm after automated pricing led to systematic overpaying, a write-down north of $500 million, and roughly 2,000 job losses. In each case, the technology did exactly what it was trained to do. The failure was upstream — in the absence of a human checking direction and reviewing output before it caused damage.
The Grant Thornton 2026 AI Impact Survey — nearly 1,000 senior business leaders — put a number on how widespread this gap still is: 46% cited AI governance or compliance failures as a leading cause of their own AI underperformance, and more than three-quarters said they lack confidence they could even pass an independent AI governance audit within 90 days. This isn’t a technology problem being reported. It’s a leadership problem, wearing a technical costume.
Who is going to make sure AI gets used well?
Not, I think, our government — at least not soon enough to matter for the decisions your business is making this year. As of mid-2026, the United States still has no comprehensive federal AI law. What exists instead is a fast-growing patchwork of state rules — Colorado, California, Texas, Illinois, and others each moving at their own pace, with different definitions of what counts as “high-risk,” different disclosure requirements, and an unresolved fight over whether federal preemption will ever arrive to unify any of it. If you’re waiting for Washington to define responsible AI use for you, you may be waiting for a while — and your competitors, your customers, and your own AI deployments won’t wait with you.
So if it’s not government, and it’s not the vendors selling the tools, it has to be the people actually using AI to run businesses. Not in theory. Not as a talking point on a keynote slide. As an ongoing, collective act of judgment.
That means people committed to AI implementations that are genuinely safe and that work within frameworks that already exist — modifying them, where they don’t yet fit, so AI serves people with the most positive outcomes and the least collateral damage. It means people with real subject-matter expertise in their own domain who are willing to sit down with people who have real expertise in different domains, and work out, together, what “done well” actually looks like before the mistake happens rather than after.
This is why HAIL exists.
HAIL — the Human + AI Leadership Council — is a private, founding-member invitation-only council built on exactly that premise: AI without human leadership is just expensive guesswork, and the people best positioned to prove otherwise are the leaders already doing the work, in conversation with each other. It’s organized around the #AIsandwich philosophy — human judgment sets the direction, human judgment reviews what comes out, and AI does the work in between. Not because AI can’t be trusted in the middle. Because the middle was never the part that required trust in the first place.
Membership in that room is private. What the room produces isn’t — more on exactly how to follow that at the end of this piece.
There’s a version of this argument that’s also just competitive reality: AI access itself is becoming a commodity. Everyone will soon have the same models, at the same price, in the same week they launch. When that happens, the only differentiator left standing is the quality of the judgment surrounding the tool — who set the direction, who reviewed the output, who was accountable for what shipped. Companies that build that judgment deliberately, now, will be the ones still standing when the tools themselves stop being a source of advantage at all.
The question, then, is a simple one.
Are you someone who wants AI to produce results that are safe, pragmatic, and genuinely superior — not results that merely look impressive in a demo? Are you someone who wants to make sure that when you use AI, or when the next generation inherits it, it’s been shaped to fit our culture and our society — rather than left to run perpendicular to it?
If that’s you, you already understand why HAIL exists. The only remaining question is whether you’re going to help build it, or watch it get built from the outside.
“Human first. Technology second.”
| How to Follow Along Founding membership in HAIL is private and by invitation — but you don’t have to wait for an invitation to see what’s coming out of it. Follow the public conversation. After every council gathering, real, anonymized insights get published for anyone to read — no membership required. Join The HAIL LinkedIn Group to follow along. Request consideration as a Founding Member. HAIL is deliberately small — twelve to fifteen leaders, each bringing a distinct professional lens. If that sounds like you, DM Steve Goldner directly to start the conversation. Know someone who belongs in this room? Share this article, or send them straight to the Group. The people who need to hear this argument most rarely go looking for it themselves. |
