I'm Nick Hern. I lead product at Thomson Reuters, I founded Kindred Tales, and I build with AI most days. This is where I show the work and write down what you can try yourself.
Featured Creation
I sell Kindred Tales gift cards at events with it. It takes live card payments on a Stripe reader, and the demo lets you run a sale yourself.
A Stripe Terminal test card lying next to a white Stripe Reader M2 on a dark surface.
A memoir business with paying customers, the AI agents that do its marketing, a point of sale checkout solution app, a data pipeline that keeps a model away from the numbers, and enterprise product at Thomson Reuters.
An AI-assisted memoir service. A parent or grandparent answers questions over a year, and the family gets a hardcover book. I founded it and run product, AI, and operations.
In production. Paying customers, books printed and shipped.
The person telling the stories usually isn't the person who bought the gift, and may never log in.
Email is one of five ways to answer. A reply to the question is enough.
The AI agents that do marketing and operations work for Kindred Tales. They research, draft, check quality, and report back. I built it, and I'm the one who signs off.
In production. It works for one company, mine.
The agents act on a real company's marketing, and what an agent reports about its own work can be wrong.
Permissions widen in stages. Every change is logged with who made it, and I approve what gets published.
A side project where every page carries figures people may rely on.
Private side project. I'm keeping the name to myself for now.
A model that writes a number can also get it wrong.
Code calculates every figure. The model only writes the prose around them.
Product leadership across self-service and account management platforms.
Enterprise work, ongoing.
Read the Thomson Reuters case studyModels, services, and platforms I've used in real projects. See the full list
Models and agents
Speech and voice
Build and host
Payments and customers
Marketing
Things you can try yourself, from a quick win at the kitchen table to a build that takes a few sessions.
Try it now
I asked AI for a game built around what my kids like. It's a real, playable word search, and you can try it right now.
The Rocky Mountain Word Hunt game. A nine-by-nine grid of letter tiles sits beside a list of eight words to find and an Easy or Tricky switch. OWL, CAVE and GOLD are found and colored in.
How I Work
On a directory site I run, every number on a page comes from code. The model writes the sentences around numbers that already exist. A separate pass checks the page against the data, and I keep reading a sample by hand.
Most of what I've learned about AI comes down to this: decide what the model isn't allowed to decide.
I went to law school, worked as a staff attorney, and moved into product in 2011. Most of the years since have been at Thomson Reuters, where I lead product for self-service and account management.
In 2022 I started Kindred Tales on the side. It has paying customers, and it's where I do most of my hands-on building.
More about meI'm happy to compare notes. Tell me what you're trying to do and where it's stuck.
