Why I Joined The Herd (Security)

I have spent a lot of time thinking about the moment I understood what I had built. Not during the build. During the build, it was just a problem to solve, a model to train, a metric to optimize. The reckoning came later, quietly, the way these things usually do. I have worked at the intersection of technology and human behavior for most of my career, and the through-line I keep returning to is this: the most powerful systems are not the ones that process the most data or run the fastest inference. They are the ones who understand people. That cut both ways then. It cuts both ways now. What follows is an honest account of what that has looked like, and what we are building because of it.

Where it all started

In 2012, I helped build a neural network used by payday lending companies. The goal was straightforward and deeply predatory: figure out the exact amount of money to loan someone so they could almost pay it back, but not quite. Just enough to trap them in a cycle of endless debt, paying back thousands of percent on a few hundred dollars.

That was over a decade ago. The world of AI has changed dramatically since then, but the lesson has never left me: machine learning can be used to hurt people.

The double-edged sword of AI

AI understands aspects of human behavior better than people do. The same capability that lets a neural network trap someone in debt can also be used to protect people. It can recognize manipulation, intervene before a bad decision is finalized, or build habits that make people harder to exploit.

That is the fundamental thesis behind Herd. We believe that just as AI can learn to exploit human psychology, it can also learn to strengthen it. The pathway is natural language: the mediums people use to communicate are the same ones bad actors use for manipulation, and the same ones through which people learn most effectively.

We believe the best way to help people is to meet them where they are. That is where they are actually vulnerable. Real language, real context, real engagement.

Why is phishing harder to spot in 2026?

Modern language models have made the situation orders of magnitude more dangerous than anything we saw in 2012, even though the parallel to my experience building payday loan machine learning models still holds.

It is worth remembering that just a few years ago, standard security awareness training encouraged employees to spot phishing emails by looking for typos. In 2026, that advice is essentially useless. Bad actors now have access to the same AI tools as everyone else, but without the compliance departments, ethics reviews, or procurement cycles that slow legitimate organizations down. Someone with a language model and bad intentions can generate thousands of unique, psychologically targeted messages in an afternoon.

This is the reality of 2026. Bad actors have the same technology and none of the friction. The gap between what is possible with AI and what most people are prepared for is growing every day.

Why did AI chatbots change how people learn?

There is a reason ChatGPT became the fastest-adopted technology in history. It was not just the capability. It was the experience. It feels human. It feels like magic. It feels limitless. People do not need to learn a new interface or translate their thoughts into search keywords. They simply communicate, and the machine responds.

This is not a gimmick. It is a fundamental shift in how humans interact with technology, and it is just the beginning. Instead of passive content consumption, users get active engagement. Instead of one-size-fits-all information, they get something that adapts to the individual in front of it.

Why does chat-based security training work?

Text-based communication should not be underestimated. There is a reason Slack achieved such rapid, widespread adoption. Text is the ideal mechanism for collaboration: ambiguity is reduced because everything is recorded, and threaded conversations allow people to stay on topic in ways that spoken communication often cannot.

Building tools that properly leverage modern communication platforms and natural language AI is still in its early stages, and the potential is far greater than most people recognize. One way to think about it: this is an opportunity for AI to actively accelerate human development. We have a chance to let AI understand what it takes to teach people effectively, to observe and refine the patterns that lead to real learning, both individually and collectively.

If the first era of AI was defined by machines learning from humans, the next phase may be defined by machines learning how humans learn. For us at Herd, this idea is foundational. As Marshall McLuhan observed, the medium is the message. Information is only as powerful as the medium carrying it allows.

Where Herd comes in

Herd exists because we have seen both sides. We know what happens when you point AI at human psychology with bad intentions. We have also seen how powerful AI tools can be when used to help people master skills and reinforce positive behaviors.

Our mission is to make good decisions easier, in the language people already speak, at the moment it actually matters. The machine knows you better than you know yourself. The question is who is holding the controls.

Join the Herd

Herd is built to empower security teams to become creators. It is the fastest, easiest way to deliver security awareness content in multiple formats that genuinely engage employees. See it for yourself.

More on why we started Herd and the people behind it — and if this is the problem you want to work on, we are hiring.