Commercial first
I set the roadmap and the budget, and I keep the engineers and the board working from the same page.
About
Chris Kedwell
Founder and CEO, Ad Magic Media
I have been building on the web since commercial websites were a novelty, and I have steered organisations through each shift since, from streaming radio to agentic AI.
I start with the commercial question. I set the budget by priority, and I define the risk before the build, particularly where AI output can be unpredictable. I keep the engineers and the board on the same page.
I founded Noopsly® so that several businesses can share one well-guarded system, rather than a string of one-off experiments. I still write the code.
The story
Each product grew out of the problem before it.
Ad Magic began in employer brand. For years we helped global employers show people what it's really like to work for them, through employer brand strategy and recruitment advertising. Everyone agreed employer brand mattered. Nobody could measure it.
So we set out to measure it. We built Vibe, a 12-criteria model of employer brand, and used AI to score organisations against it. That work became Noopsly® Evaluation.
We quickly learned that one model isn't enough. A single AI model can be confidently wrong, and you can't see when it is. So we built a quorum: several models from different providers, each scoring independently, with the answer taken from what they agree on. It's more reliable, and every result can be audited.
Employers also needed to get their message out at scale. Talent Avatars® turn job descriptions into recruitment videos, and can interview candidates too.
Running all of this needed AI that could remember, act and answer to us. Rather than rent it, we built it: our own memory layer, Purple Hub, and our own agent platform, Noopsly® Agent.
Today that stack runs real businesses. Canada's Entrepreneur and Debs & The Horse run on Noopsly® Agent in production.
And we keep going. Quplet is our research into what comes after today's AI.
Background
Hands-on
I design and build across AWS, Go and Python, including Purple Hub and Noopsly® Agent.
The LLM quorum
Several models have to agree before the output is trusted, and every service can fall back to a second provider.