The traditional signals VCs use to evaluate early-stage companies have fundamentally changed with AI. At Bowery Capital, we've had to rethink how we assess opportunities in a post-AI world. Curious, what else would you add? Email me at loren.straub@bowerycap.com.
What Changed: From Technical Execution to Defensibility
Product sophistication is no longer a reliable signal. With tools like Replit and countless emerging AI development platforms, a polished MVP with intuitive design can be built in hours rather than months. When a founder walks into our office without a product today, it suggests either unfamiliarity with modern development tools or minimal time investment, neither of which inspire confidence as investors.
This democratization has also flipped our evaluation criteria. We now question founders who build completely from scratch rather than leveraging AI-powered tools and open-source frameworks. The smart money focuses innovation precisely where it creates differentiation, not on reinventing foundational components.
Domain expertise + AI-native has become the new moat. Since anyone can build, the critical question is: what should you build, and why does it matter? Deep domain expertise, having lived the pain point daily, is essential. These experts no longer need technical co-founders to translate their vision; they can execute directly while maintaining the purity of their insights.
For vertical AI opportunities especially, we need demonstrable expertise that goes beyond surface-level understanding. The founder must possess insights that couldn't be replicated by an LLM or weekend research session. This doesn’t necessarily mean you’ve spent decades in the industry, it means you’ve used all the tools at your disposal to become an expert in the category. With GPT Deep Research, Manus and other tools, you can become an expert in months vs. years.
Defensibility requires entirely new frameworks. AI is simultaneously creating new markets while making existing ones more contestable. Traditional "first-mover advantage" thinking breaks down when fast followers can build competitive products in weeks rather than years. The old moats, proprietary technology, complex integrations, or sophisticated UX, can now be replicated rapidly using AI development tools.
This means the moat has to come from something deeper: proprietary data relationships that improve with scale, regulatory approval that takes years to achieve, network effects where value compounds with each user, or embedded workflows that become integral to operations. Harmonyze is an example of a startup using a compounding data advantage to ensure new AI entrants will struggle to compete against them on performance. Harmonyze sells a vertical AI platform for franchisors; having been first to this category, they had a headstart and worked with their early customers to train their system to extract key data points from proprietary franchise documents with a higher degree of accuracy than any new entrant would be capable of. This advantage stems from Harmonyze’s access to non-public franchise operating agreements, which are data rich but difficult to obtain and can vary significantly from franchisee to franchisee.
Compounding data advantages that come from training on non-public industry documents and workflows can help early entrants into a given vertical stay in the lead over the long run. We now explicitly ask founders: "If a well-funded competitor spent six months building your exact product using AI tools, what would prevent them from succeeding?" If the answer is "we'd build faster" or "our features are better," that's no longer sufficient.
The definition of shipping with "high velocity" has changed. Our best performing businesses are shipping product at a speed we’ve never seen before. They are deeply listening to customer feedback and building out new features in functionality within 24-48 hours. This is the new expectation we have for founders and believe this short time to “wow moments” will ultimately be the defensibility for the next wave of businesses. Bias to action > perfection.
What We're Investing In: Systems of Action Over Systems of Record
Our investment thesis has crystallized around several key areas:
Vertical AI with proprietary data flywheels. In vertical SaaS, data isn't just a byproduct, it's a moat that deepens with each customer interaction. Companies that harness this effect to deliver increasingly intelligent solutions will command sustainable advantages.
Systems of action that move beyond passive record-keeping to enable humans, AI-assisted humans, and autonomous agents to act on data and trigger workflows. These systems of action represent the evolution beyond traditional systems of record.
Human augmentation in complex, regulated industries where compliance creates natural barriers and high-stakes decisions make AI augmentation particularly valuable. We're drawn to solutions tackling frequent information handoffs, preserving context between physician shifts, maintaining customer service continuity, or ensuring critical information never falls through cracks.
The pattern: AI excels as the "always-present employee" that enhances rather than replaces human capabilities. Rulebase, one of our more recent additions to the portfolio, is emblematic of how we see opportunity for AI to augment human judgment. Rulebase’s platform gives CX teams at fintechs and financial institutions the ability to answer any customer inquiry with perfect recall and in line with all applicable compliance requirements. Given the sensitive nature of these conversations, the fintechs that Rulebase supports still want a human to handle the discussion, but now that representative has been supercharged by AI and can be a “10x rep” by allowing Rulebase to ride along on every call and provide real-time insights and suggestions for even the most complex customer interactions.
The Bottom Line
The most successful investors won't chase AI as a category, they'll understand how AI is reshaping every category. Companies defining the next decade aren't just using AI; they're being built differently because of it.
For founders: demonstrate deep domain expertise, leverage AI tools efficiently, and design for AI-era capabilities rather than automating existing workflows.
For investors: develop new frameworks for this landscape. Many traditional signals have become noise. The future belongs to those who recognize that shift.
If you liked “Rethinking VC Evaluation” and want to read more content from the Bowery Capital Team, follow us on Substack to get our latest insights delivered directly to your inbox.
