Some companies can run on a playbook. Hire a strong operator. Install a process. Scale the motion.
Genomics does not work that way.
Dr. Brandon Colby made that clear in our conversation. “When you are a CEO of a company that’s dealing with very complex data, very scientific data,” he said, “it takes a really intricate, deep understanding of that technology thing from the very top… in order to understand what products to focus on [and] what products not to focus on.”
That is the first uncomfortable truth for CEOs building at the edge of science. The job is not just leadership. It is leadership plus technical judgment at a depth most industries never require.
Brandon is a medical geneticist, physician, and founder of sequencing.com, a direct to consumer platform for whole genome sequencing. He has spent two decades translating clinical science into consumer products. His frustration started in the exam room.
When Research Stays Stuck
Brandon did not leave medicine because he disliked patients. He left because he could see a larger bottleneck.
He watched new genetics research come out “on really a weekly basis,” associating genes with disease, and saw that it was not making its way into practice. “I saw all this research, but it was not being put into the hands of myself and other practicing doctors,” he said.
That gap was not a minor inefficiency. It shaped his entire career choice.
He loved practicing medicine. “Treating patients… was extremely fulfilling,” he told me. Then he said the line that explains the pivot. “I was not gonna be able to change the world… That was only gonna be possible through business.”
You can hear the clarity in that. He wanted scale. He wanted research to become usable.
Healthcare adoption is slow for reasons that sometimes make sense, safety matters. Brandon agreed. He also pointed to another drag that CEOs should recognize. “It was a lack of understanding of how this information could be used,” he said. People were building libraries of research, then stopping. “There were no grants… about how to translate that into clinical care.”
Doctors expected someone else to do the translation. Brandon decided to be the someone else.
The CEO Blind Spot In Emerging Tech
I asked Brandon what he sees other CEOs get wrong when they try to grow companies built on complex, emerging technology. He went straight at the most common mistake. CEOs assume they can lead deep tech the same way they led anything else.
Brandon does not buy it. “It’s not where you can just go and bring in a CEO that has a playbook from another industry,” he said. “Instead… it takes a CEO that has a very intricate knowledge of that technology.”
This is not about ego. It is about decision quality.
When the product is built on scientific nuance and massive data, small mistakes compound. You need someone who understands the limitations, what has failed before, what has worked, and why. Brandon described how his early decisions were driven by understanding “its limitations… how that technology has been used in the past and hasn’t been successful, what has been successful.”
That depth shapes everything.
- Strategy
- Product scope
- Hiring
- What you deliberately do not build
If you cannot evaluate tradeoffs at that level, you will chase shiny features and miss the real constraints.
Clinical Context Is A Force Multiplier
Brandon also has something many technologists do not. He lived the problem in the clinic.
He understood the workflow and the education gap. “Most genetic education is about two to three weeks out of four years in medical school,” he said. That means most physicians are not equipped to use genomics without support.
He also described the culture Mount Sinai ingrained in him. Learn the current state of medicine, then challenge it. “It was our duty as physicians… to ask questions, to disrupt the status quo in order to move medicine forward,” he said.
That matters because genomics is not just a data problem. It is a behavior change problem inside a slow moving system.
Scaling When You Are The Expert
Deep tech CEOs face a specific trap. You are the expert, so you stay in everything. Then the company hits a size where that stops working.
Brandon’s approach is straightforward and humbling. “The most important thing that I’ve learned… is to hire people that are smarter than you,” he said. Not vaguely smart. Smarter in specific domains.
He described handing off parts of the business to people with deeper expertise. “That is definitely humbling,” he said. He also described what happens when you build that into the culture. The entire team levels up.
He went further. Humility is an ethos at sequencing.com. “No matter how good we think we are, there are other people out there that are smarter,” he said. “We want to be humbled by them, and learn from them.”
That mindset is not soft. It is a scaling mechanism.
AI Is Useful, But Not Where You Think
Brandon is “very bullish” on AI. He also called out the hype.
He described the media narrative: engineering teams cut 90%, five people ship five years of work in a day. Then he gave the reality from his own company. “It’s not there yet,” he said. “A lot of that is overblown as hype.”
That does not mean AI is useless. It means you need to put it in the right place.
Operationally, he is using AI in customer success. Not to replace humans, but to multiply them. “They all now run different agents and different AI agents report back to them,” he said, allowing each person to accomplish far more. That matters for another reason. It lets them keep support onshore. “Everyone being based in the U.S.,” he said, and AI made it easier to scale without outsourcing.
This is the right model for most CEOs.
- Humans own the customer
- AI increases throughput
- The company keeps quality and context
Three Billion Data Points And The AI Trap
Whole genome sequencing brings a different scale of complexity. Brandon put numbers on it.
“One person” is “about three billion data points,” he said. The output from the sequencer is roughly 70 gigabytes, and the full set of files they work with can be “about 100 gigabytes of data” per person.
That volume makes people assume AI should do the analysis. Brandon says that is exactly the mistake.
He was blunt. AI is not capable of accurate genetic analysis right now. He described what happens when people upload consumer genetics files into ChatGPT. “It does perform an analysis,” he said, “but… it’s primarily incorrect.”
The reason is not just training volume. It is nuance and translation. “It hasn’t been trained on the human genome, on the nuances of interpretation,” he said. Even identifier systems change over time. AI may read an older paper and misapply IDs to current systems. Their algorithms know how to translate those changes. AI does not.
So Brandon made a disciplined product decision. “None of our analysis is based on AI,” he said.
Then he described where AI is outstanding. Interpretation and communication.
AI can take validated results and present them in the right form for different users.
- A physician wants results in a clinical format
- A layperson worried about brain health wants a narrative explanation
- Different customers need different depth and framing
“The translation of the results into many different presentations, many different reports and discussions,” he said, “the AI is absolutely outstanding at.”
That is an important product lesson. Use AI where it is strong. Keep it out of the core where mistakes are dangerous.
What You Should Take From This
If you are building in frontier tech, leadership is not enough. You need credible technical judgment at the top or you will make the wrong bets.
Brandon’s model is worth borrowing.
- Deep CEO understanding guides focus and restraint
- Hire people smarter than you and let them own domains
- Use AI to multiply teams, not to replace accountability
- Keep AI away from analysis where correctness is not guaranteed
- Use AI to translate complex results into useful, human language
That is how you scale science into something people can use. Listen to the full episode of The Scaling SEO here.
I am Glenn Gow. I coach CEOs. If you are building with complex data, dealing with AI hype, and trying to decide what the CEO must understand personally versus what you can delegate, I can help you build that operating model.
