Andrew Gershfeld, General Partner at Flint Capital – Interview Series

Andrew Gershfeld, General Partner at Flint Capital, leads the firm’s international business development efforts and serves on the boards of JobToday, Cream Finance, Yva, and Flo.Health. He previously held the role of Chief Investment Officer at ABRT Venture Fund, where he focused on early-stage deals and co-investments with major firms. Before entering venture capital, he built and exited his own company, GlavSkidka, gaining firsthand entrepreneurial insight that now informs his expertise across consumer internet, eCommerce, B2B SaaS, and virtualization sectors.

Flint Capital is a Boston-based global venture firm that backs early-stage software companies across the U.S., Europe, and Israel, with a focus on categories such as enterprise SaaS, AI, cybersecurity, fintech, IT infrastructure, and consumer internet. The firm helps founders scale into the U.S. market and typically invests from pre-A through Series B, leading rounds with checks ranging from $1–5M and following on up to $10M. With notable successes such as WalkMe, Socure, and Flo.Health, Flint Capital has built a reputation for supporting high-growth startups and has been recognized multiple times as a founder-friendly fund.

You transitioned from building and selling your own company to becoming a VC and eventually a partner at Flint Capital — what originally pulled you toward investing, and how did that shift influence the way you evaluate and support founders today?

Curiosity pulled me in. I had bootstrapped my own company while watching other founders raise venture capital — funding collateralized by nothing but vision and energy. I kept wondering: Is this charity, or something deeper?

After my exit, I became an investor to find out. And here’s what I learned:

Money is the same in everybody’s hands. The difference is in the hands that hold it. Some VCs become rocket fuel. Others become a drag on the company.

Before I became a VC, I assumed it was only about capital. The biggest surprise was realizing how much impact comes from everything around the capital: asking the right question at the right moment, listening without judgment when a founder needs to think out loud, making the right introduction that triggers a chain reaction, or even helping close a critical hire by taking a candidate to lunch and showing them the opportunity of their lifetime.

And sometimes, the job is colder — pouring water on the team when product-market fit isn’t there and the money clock is ticking.

My founder years taught me three things I carry into every investment:

First, respect for the grit of bootstrapping. Founders who delay raising as long as possible build a different level of resilience.

Second, a deep understanding of the profound loneliness of building something frontier. It takes immense self-confidence to keep moving forward and inspire others to follow you into a territory where no one has gone before.

Third, that the team defines everything. Solopreneurship is a real trend, and we may see unicorns built by single individuals. But in my experience, outcomes are determined by the team more than anything else.

The hardest part of the job is balancing support with directness when a portfolio company is drifting the wrong way. Investors fear being misunderstood, but founders almost always return later with appreciation. Those conversations help them reset, ground themselves, and recalibrate for what’s next.

There’s no formula for this. The only constant is trust, and trust requires vulnerability from both sides. Founders who always appear as alphas and insist on projecting invincibility are missing something critical. Real connection happens in moments of vulnerability and honesty, and that’s where founders build deeper, more sustainable roots.

When you assess AI founders, what signals tell you they can build a durable company — and how do you balance technical excellence against deep domain expertise in high-trust sectors like legaltech, cybersecurity, and healthcare?

Most VCs optimize for durability. I optimize for anti-fragility.

Durable companies withstand shocks. Anti-fragile companies get stronger because of them. Durability is defense, while anti-fragility is offense. And in the AI age, adaptation speed beats resilience every single time.

When I meet founders, I don’t look for people who can endure stress. I look for people who transform stress into new strengths and opportunities. That’s what forges great companies. Identifying this quality is complex, and to do it we examine how they’ve dealt with uncertainty historically. It is this anti-fragility that is the signal I’m hunting for.

As for technical excellence versus domain expertise: In AI, domain expertise wins. Every time.

For early-stage companies, domain expertise and the ability to sell and distribute matter more than technical brilliance. We still want technical talent in the form of people who see the frontier and can chart a course to get there correctly. But they don’t need to be master craftsmen. They need to ship fast.

I’d rather back a lumberjack who can deliver product-market fit in three strikes than a wood carver spending years crafting the perfect architecture. Speed to PMF is everything, and domain experts know exactly where to swing the axe.

The ideal founder? A domain expert who has directly felt the pain they’re solving, someone who knows the terrain and moves urgently.

And for team structure, give me two founders who trust each other completely, understand their lanes, and complement each other. That configuration wins far more often than any specific skill set.

Proprietary data, workflow integration, and defensibility are becoming mandatory for early AI companies — which of these do you view as the strongest moat, and why?

Models are commoditizing like CPUs. The moat is in the data and the workflow. They’re interdependent, one strengthens the other.

Models are the brain, and they’re getting cheaper every quarter. But the output you generate depends entirely on the data you use to train them, form requests, and augment answers. To collect that data, you need touchpoints with users and workflows. The workflow produces data. The data strengthens the workflow. It becomes a flywheel.

But where it gets interesting is that seamless integration, more than being a retention strategy, involves creating a completely different paradigm.

There’s a massive difference between copy-pasting into ChatGPT, waiting, then copy-pasting back versus an AI that lives in your workflow, understands context, and executes without you ever leaving the environment. The former is a tool. The latter is a co-worker.

The future is a new interface entirely — a Semantic OS, an AI browser. A single environment where intent and execution happen together with no context switching.

We’re actively investing in this paradigm shift. LeoAI, one of our portfolio companies, augments engineering teams with a virtual engineer embedded in their environment. It understands project context and augments design the same way Cursor understands a codebase. It schedules. It commits to the repository. It’s a team member.

The litmus test is simple: Can users stay in the environment from start to finish without leaving? If yes, you’re building the future. If not, you’re building another API wrapper.

Many AI companies struggle to convert pilots or proof-of-concepts into repeatable revenue — what evidence convinces you that a startup is moving beyond experimentation toward real product-market fit?

In the age of AI, one metric matters above everything else: engagement.

Not revenue, contracts, or pilots. How many eyeballs are captured and how much time do they spend? How much work is completed inside the AI environment? That’s it.

We’re in a war for attention, and attention is the leading indicator of everything else.

In early stages, retention data isn’t always available. But you can spot patterns. If the product is genuinely helpful and rich enough for early adopters, you’ll see stabilizing churn within weeks. That’s the signal that separates stickiness from novelty.

The metric that predicts everything, though, is how much work gets done inside the AI environment. If users are reaching their goals and actually completing tasks, solving problems, and producing outcomes, you’re looking at real product-market fit.

This is true whether it’s consumer or enterprise. AI automates human work. Humans get more done with AI. So if you see constant utilization, constant activity, and substantial outcomes, you’re watching PMF emerge.

Enterprise is slow. B2B contracts take time and require proof of concepts, lengthy sales cycles, and ultimately, integration hell. But when sophisticated buyers sign contracts, that’s validation. That’s the moment you know you’ve moved beyond experimentation.

Consumer or enterprise, the truth is the same: engagement predicts everything.

Flint Capital recently backed AI companies across legaltech, cybersecurity, and mechanical engineering — what common threads do you see among the startups winning in these diverse verticals?

The winning companies aren’t building integrations. They’re building co-workers.

We’re seeing the emergence of a new operational layer, which is AI embedded in the workflow as a true peer. This is happening across legaltech, security, and mechanical engineering, among many other fields. AI now understands context. It doesn’t rest. It works in parallel. It never leaves the environment.

This isn’t Microsoft Word with a ChatGPT plugin. It’s a fundamentally different paradigm. And here’s how it plays out across verticals:

Legaltech is the obvious win. Legal language is code, legislative code. LLMs are built for this. Same with cybersecurity. You’re analyzing code, structure, and dependencies. It’s all abstractions describable in words.

Mechanical engineering is the hard one. These aren’t language models but mechanical ones. They need to understand spatial relationships, physics, and material science. It is a completely different language, and it is much harder to implement.

However, the endpoint is identical across all three: an environment where the professional can complete their work without switching context.

Legaltech is fast. Mechanical engineering is slow. But they’re all heading to the same destination.

As an investor, what signals help you spot unicorns early — and how do you position yourself to enter top-tier deals before they become competitive?

You can’t spot unicorns. You can only be there before everyone else is looking.

The game isn’t pattern recognition. It’s proximity. Be in the founder’s life early, before the deal heats up and the brand-name VCs show up. That’s the only way for small funds to win.

Large funds are built to deploy billions. To do that, they need maximum visibility and dealflow. They can’t miss opportunities, so they win on brand and attention.

Small funds can’t compete with that. We can’t out-brand Sequoia. We can’t out-market a16z.

So we play a different game: generational craftsmanship. Knowledge and reputation passed from one generation of partners to the next. A deep understanding of current trends, founder culture, and language shifts. We stay close to the points where ideas emerge and founders start collaborating.

Our game plan has three parts:

One: Portfolio founder referrals. No one knows you better than the people you’ve worked with. Our best dealflow comes from founders we’ve backed.

Two: Entrepreneur LPs. We’ve built a diverse base of successful founders who became LPs and formed an active community. They spot opportunities under the radar, early, before top-tier investors notice.

Three: Community as competitive advantage. A community needs something in common. We fundraise from tech founders who benefit from the brilliance of other participants. We facilitate activities, meetings, and introductions. We keep the engine running.

One more thing: celebrity investors building funds off social capital. That can work, but only if they share the playbook of converting that social wave into business opportunities. If they’re just lending their name, it’s hollow. If they’re teaching the methodology, that’s real value.

Big funds win on brand. We win on relationships. That’s the game.

What red flags make you walk away from an AI opportunity even when the technology looks promising on the surface?

The moment the next model ships, your features get absorbed. No one pays you anymore. There’s tiny margins, a short lifespan, and zero defensibility.

That’s an instant red flag.

Beyond that, there are other factors such as team dysfunction, broken unit economics, or structural churn embedded in the business model. These are  existential problems more than execution ones, which means you can’t fix them by working harder. They’re baked into the DNA of the business.

When you see that, walk away, regardless of how impressive the demo looks.

What types of AI companies do you believe are best positioned for outsized growth over the next decade — workflow-embedded tools, vertical-specific solutions, infrastructure, or something else entirely?

A: Infrastructure will be huge, but capped. Commoditization crushes margins. We don’t see trillion-dollar companies producing commoditized services.

The real winners will be companies that control the user surface.

Whoever owns the interface, the environment where humans work, will continuously outgrow everyone else. They’ll be sustainable and have the competitive advantage.

But the biggest transformation will come from something else: The Windows 95 moment for AI.

When the desktop paradigm was introduced, it changed everything. We’re about to see something similar for human-AI interaction. A new interface. A new paradigm. And it will transform the entire industry.

The disappointing part is that this breakthrough won’t come from a startup, but from companies we already know. You need a deep understanding of behavioral patterns to pull this off. How users interact. What incentives drive them. What problems models face. How memory works. How to maintain context without losing it.

Newcomers would have to figure this out from scratch. Existing players already know it.

So we know the names. We just don’t know who wins.

Startups will win verticals. Incumbents will win the platform shift.

You’ve helped many companies expand from Europe and Israel into the U.S. — what differentiates AI startups that successfully scale into the U.S. market from those that struggle?

Proximity to the customer is everything.

The companies that win are physically present in the U.S. spending time with customers, design partners, closing deals, and pushing them through the pipeline to production. Those that prioritize go-to-market and pivot quickly win. Those that try to scale remotely struggle.

Engineering can be anywhere, but sales and product must be in the U.S. from day one.

You can save costs by keeping R&D in Israel or Europe. But if your front office, including sales, product, and customer-facing teams is remote, you will not succeed.

Product-led growth changes the equation. If you’re targeting SMBs or consumers with a PLG motion, you can run marketing remotely as long as your team understands U.S. funnels, PR, and growth tactics.

But for mid-market to enterprise? No way. You have to be there.

Trying to save money by staying remote is a false economy. You’ll spend more fixing the damage later.

For founders building in regulated or high-trust sectors, what advice would you give them to accelerate traction, deepen customer trust, and raise capital more effectively?

Fail fast, move fast, and ship fast.

Don’t wait for the perfect product. Don’t wait for the market to be ready. Don’t wait to hire the perfect person. Move now. If you do this, you’ll iterate and learn faster, and understand your customer’s pain points before the money runs out.

Regulated AI is expensive. Seed rounds are massive because teams cost a lot, the technical challenges are hard, compliance is overhead, and you’re facing bigger clients in corporate environments. All of that requires more capital and more investment in the product.

The faster you move, the better your chances of survival.

In regulated AI, there’s no prize for perfection. There’s only survival for speed.

So, again: Fail fast. Learn faster. Ship fastest.

Thank you for your insights on investing, readers who wish to learn more should visit Flint Capital.