What does Adam Roozen do?

Adam Roozen builds technology systems and trades his own capital. He is CEO of Isotropic Solutions, which builds governed agent workforces for businesses, enterprises and national institutions. He also runs business automation experiments testing how much can operate without people, and takes a small number of advisory engagements each year.

What is Adam Roozen's background in trading?

Adam has traded his own capital for ten years, and it is his primary income. He started in forex around 2016, added crypto in 2018, and crypto became the main focus from 2021, with equities and index futures alongside it. He trades theses rather than tickers, and has built software around his own research process. He trades only his own capital. He does not manage outside money, is not a financial adviser or fund manager, and publishes no performance figures.

What is Adam Roozen's background in digital technology?

Adam has worked through every consumer technology wave since the mid-nineties. He built transactional websites in high school, before payment processors existed, and had to write his own card encryption. He joined Walmart as a programmer analyst during the early dot-com build-out, ran e-commerce marketing at Sam's Club through the mobile wave, scaled a global commerce agency through the social and cloud waves, worked in tokenized and blockchain systems, and now builds autonomous agent systems.

What is Isotropic Solutions?

Isotropic Solutions was founded by Seran Devaraj, and Adam Roozen joined afterward as CEO. The company builds governed agent workforces: agents do operational work, a person signs off, and every decision keeps its evidence. Autonomy is earned a rung at a time, from drafting, to recommending, to acting under review, to running inside written limits. The second half of the work is preparing businesses to be read and transacted with by other companies' AI systems.

How is Adam different from a traditional technology consultant?

Adam runs the systems he would otherwise be advising about. Two of his companies operate with no employees, his trading runs on software he wrote, and Isotropic Solutions ships agent systems into banks and commodity traders. He has also spent nearly 30 years operating: building early e-commerce at Walmart, running e-commerce marketing for a $700M to $1B channel at Sam's Club, and co-founding a 200-person agency.

What is Adam Roozen's background in e-commerce?

Adam led e-commerce marketing and operations at Sam's Club (Walmart) while the channel grew from $700M to $1B in annual revenue. He then co-founded Echidna and scaled it to more than 200 people, serving blue-chip retailers including Kohl's, Michael Kors, Toys 'r Us, Lowe's, Cracker Barrel, Aveda, Jelly Belly, Wolters Kluwer, Dubai Duty Free, Liverpool in Mexico and Mr. Price in South Africa.

Who does Adam work with?

The work spans four sizes. One person, in the zero-employee companies he runs himself. Owner-operated businesses between $5M and $50M, through W3 Strategic Management Services. Enterprises, such as ETG, a global agricultural commodity trader turning over $8B across 250 business units. And national institutions, including the Central Bank of Oman and Vietnam International Bank.

How do I start working with Adam?

Adam takes a small number of advisory engagements each year, and works directly with whoever is accountable rather than through a team of associates. When an owner-operator needs the cash more than he needs the invoice, he will take revenue share instead of a fee. The best first step is a short note through the contact page describing what you are up against.

Hidden Meta: The Routing Layer

Route most of your traffic to small models and the bill falls 40 to 85%. The edge moved out of the model and into whatever decides which model to call.

A Forbes headline from June 25 would have sounded absurd a year and a half ago. Small language models beat frontier AI on cost, on speed, and on accuracy.

Not on some narrow benchmark. On the things buyers actually pay for.

I've run my own stack this way for a while, so the finding didn't surprise me. What surprises me is how few companies have restructured around it.

Route 70% of traffic to a small model and 30% to a frontier model. The input-token bill drops about two thirds. Push it to 80/20 and you approach 79%. Teams that tune the routing layer report 40 to 85% savings with no visible quality loss.

Getting a Baseline

A frontier model is trained to be good at everything. So what are you buying on a call that tags a support ticket? Capability you never use.

Classification runs fine without it. So does extraction, summarization, reformatting, tagging and validation.

That's most of what a production system does all day.

Llama 3.1 8B, Phi-3 Mini, Mistral 7B and Haiku handle that work for a fraction of the price. They also answer faster, which matters more than people expect once a workflow chains ten calls together.

The Shape That Works

One frontier model sits at the coordination layer. It reads the task, breaks it into pieces, and hands each piece to a worker running something cheap.

It only comes back when a worker hits something it can't resolve.

Expensive tokens go where judgment happens. Everything else runs at commodity prices. If that sounds like how you'd staff a team of people, it should.

So Where's the Edge?

Everybody has the same models. Nobody gets an edge from picking Opus over Gemini.

The edge is in the routing logic, the verification loops, the small models tuned on your own data, and the evals that tell you when a route quietly started degrading.

That's the part a competitor can't buy. It's also the part most AI budgets underfund, because it doesn't demo well to a board.

The Trap

A badly tuned router is worse than no router.

Send a hard problem to a 7B model and you don't get a cheaper answer. You get a confidently wrong one, and wrong answers cost more downstream than tokens ever will.

The real work is in escalation – a worker has to know when it's out of its depth, and it has to say so instead of guessing. Most teams skip that part, then blame the small model.

A Question for Your Vendor

What percentage of your calls hit a frontier model?

If the architecture is one big model answering everything, you're paying frontier prices to classify support tickets. If they can't tell you the number, that's your answer.

Wrap-up

Here's why I care beyond the invoice.

If most production tokens flow through small models on cheap hardware, then demand for the biggest training runs and demand for inference capacity stop moving together.

Right now the market prices them as one thing.

I don't think that holds.