Glenn Gow, The Scaling Executive Coach and host of The Scaling Executive Podcast, talks with Gregory Shepard, CEO of Startup Science and General Partner at BOSS Capital Partners, about why structure, not talent, decides which founders scale and which stall.
When Gregory Shepard sold his companies and set out to understand why 90% of startups fail, he expected to find a talent problem. He found a structure problem instead. After five years and 1,200 founder interviews, Shepard’s research shows that 47.1% of founders fail within their first 18 months, and the failure almost always traces back to one decision: skipping the vision phase where a founder defines the customer, the buyer, the competitive landscape, and the case for the product before building it. Founders who build structure into that early phase, and later into how they delegate and how they adopt AI, are the ones who convert product-market fit into a company that scales.
This episode is for CEOs who have survived the first 18 months and now need a repeatable structure for delegation and AI adoption instead of relying on founder instinct.
Key Takeaways
- Shepard’s five-year research project of 1,200 founder interviews found that 47.1% of founders fail within the first 18 months, and most later-stage failures trace back to decisions made in that window.
- Shepard’s research shows founders who rush past the vision phase to start building product face a “wall closing in” effect: the longer they wait to pivot, the more their product, investors, and customers solidify around the wrong direction.
- Delegation fails when a founder trains a new hire once, then keeps checking in as if the hire never happened. Employees consistently tell Shepard the same thing: train me, then get out of the way.
- Startup Science requires every employee to complete AI training specific to their function and document one use case. The result: a job that took six people in marketing dropped to one, and a job that took ten people in engineering dropped to one.
- Shepard’s framework for AI adoption separates the roles clearly: data becomes information once it is structured, and information becomes wisdom once it is acted on. AI handles the first two steps. People still supply the third.
Skipping the Vision Phase Causes 47% of Founders to Fail Within 18 Months
Gregory Shepard’s research began with a simple question: why does an industry tolerate a 90% failure rate that no other industry would accept? What he found first was fragmentation. Unlike music, retail, or affiliate marketing, entrepreneurship has no consolidated standards that founders can learn from. Every founder starts from zero.
The deeper finding was more specific. Founders who fail are not failing because they lack a good idea. They are failing because they move into product before they finish defining the world their product will compete in.
“I think that the biggest thing was that basically when founders start businesses, they don’t put enough time into the vision phase, that very beginning part,” Shepard said. Founders skip past the customer detail, the buyer profile, the case studies, and the competitive landscape because they believe they already understand it. Shepard calls this the North Star, and founders who bypass it hit a wall later: the farther a company moves down its life cycle, the harder it becomes to pivot without spooking investors and customers who have already built expectations around the original direction.
That early gap produces the number Shepard’s research keeps returning to: 47.1% of founders fail in the first 18 months, and nearly all the failures in the following five years trace back to decisions made in that same window. Shepard names the specific blind spot behind this pattern directly: founders see their product through their own optimistic lens, not through the lens of the customers, competitors, and acquirers who will ultimately decide whether it works.
Delegation Is the Hardest Structural Habit for Scaling Founders to Break
Once a company survives its first 18 months, the structural gap shifts from vision to delegation. Shepard, who spent 15 years as CEO of Affiliate Traction through the dot-com collapse and a global recession, calls delegation the hardest habit for a founder to break as they move from building everything themselves to leading a team.
The problem is not a skills gap. It is a trust gap. Shepard is direct about why it forms: a founder who has built every function of the business personally defaults to believing they can still do each task better than the person they just hired. When a founder finally hires someone, the instinct is to keep checking in, the way a new parent calls the babysitter every twenty minutes. That oversight is what breaks the delegation, not the hire itself.
Shepard’s structural fix is direct: train the person completely, then step back. He described the pattern from the employee’s side, based on what he has observed repeatedly across his founder interviews: “The people are like, listen, train me, and then let me do my job and get out of the way.” A founder who keeps hovering after training is complete ends up frustrating the person they hired and doing the job twice, once themselves and once through supervision.
CEOs Should Structure AI Adoption Around Data, Not Just Tools
The same structural discipline that applies to vision and delegation now applies to how CEOs bring AI into their organization. Shepard’s first objection is to the framing itself: he pushes back on founders who describe themselves as having “an AI startup,” since nearly every company today runs on AI software the same way nearly every company runs on software generally. The structural question is not whether to use AI. It is how to structure its use inside each function.
At Startup Science, Shepard required every employee to complete training on how AI applies to their specific role, watch instructional material, and write up a use case for how AI could create efficiency in their function. The results were direct: a job that previously required six people in marketing dropped to one person, and a job that previously required ten people in engineering dropped to one person.
Shepard’s framework for where AI fits and where it does not fit is a three-step hierarchy. “I think of data as the raw form of information, and then data becomes information when it becomes structured. And then information, after it’s executed, becomes wisdom,” he said. AI handles the first two steps well. It cannot yet replace the third. “That wisdom component you need people for still. But the data component and the structuring component you can do with AI,” Shepard said, which is why he still leans on mentors, coaches, and advisors to add judgment once the AI has done the structuring work.
The Structure Framework From This Conversation
| Principle | What it means in practice | Named evidence from this interview |
| Structure beats improvisation | Founders who spend the vision phase defining the customer, buyer, and competitive landscape before building product avoid the “wall closing in” effect that makes later pivots costly | Shepard’s five-year, 1,200-interview research project found vision-phase neglect behind most of the 47.1% of founders who fail within 18 months |
| Delegation requires training, not just trust | Founders must fully train a new hire, then step back completely, instead of continuing to check in as if the hire never happened | Shepard identifies hovering after training as the top delegation failure he traced across his five-year, 1,200-founder research study, a conclusion he also draws from his own 15 years leading Affiliate Traction through the dot-com collapse and a global recession |
| AI structures data, people supply wisdom | CEOs get the most value from AI when they use it to turn raw data into structured information, then rely on mentors and advisors for the judgment AI cannot yet provide | Requiring every Startup Science employee to complete function-specific AI training reduced a six-person marketing job to one person and a ten-person engineering job to one person |
Quotes from This Episode
- “It’s so hard when you’re doing everything yourself and you think you can do everything better than everybody else.” — Gregory Shepard, CEO, Startup Science
- “The walls sort of close in.” — Gregory Shepard, CEO, Startup Science
- “I think of data as the raw form of information, and then data becomes information when it becomes structured. And then information, after it’s executed, becomes wisdom.” — Gregory Shepard, CEO, Startup Science
Frequently Asked Questions
How should a CEO structure AI adoption inside their company?
Gregory Shepard’s approach is to require every employee to complete AI training specific to their own function and document a use case for how it applies to their role, rather than adopting AI tools generically. At Startup Science, this structure reduced a six-person marketing job to one person and a ten-person engineering job to one person. Shepard’s framework treats AI as the tool for turning raw data into structured information, while reserving judgment and wisdom for experienced people.
Why do most startups fail in the first 18 months?
Shepard’s five-year research project of 1,200 founder interviews found that 47.1% of founders fail within their first 18 months, and most later failures trace back to decisions made in that same window. The root cause is skipping the vision phase: founders move into building product before they fully define their customer, buyer, and competitive landscape, which makes later pivots costly once investors and customers have already formed expectations.
Why is delegation so hard for founders who are scaling their company?
Delegation fails when a founder trains a new hire once and then continues checking in as if the hire never happened, which frustrates the person they hired and undermines the point of hiring them. Shepard’s fix is structural: train the person completely, then step back entirely. Founders who scale successfully treat that handoff as complete rather than as an ongoing test.
CEOs Work with Glenn Gow to Scale Their Companies by Scaling Structure First
Glenn Gow is The Scaling Executive Coach. He coaches ambitious executives into the CEO seat and CEOs into successful exits, drawing on 25 years as a CEO and 5 years in venture capital. If this conversation was useful, you can apply for executive coaching with Glenn Gow or apply to be a guest on The Scaling Executive Podcast.
