AI Made Building Startups Too Easy: Winning Customers Is The Real Challenge

1 hour ago 4

Krish Ramineni is CEO & Co-founder at Fireflies.ai, an AI teammate for meetings used by people at 75% of Fortune 500 companies.

A decade ago, starting a software company was expensive. You needed engineers, cloud hosting, infrastructure, design and months of development just to ship something mediocre. Most founders never even got past the prototype stage because the barrier to entry was too high. Now, a single person can build an app over a weekend with AI.

That changes everything. Vibe coding is creating a completely different startup environment. Founders are shipping products without engineering backgrounds. Designers are launching SaaS products. College students are building AI agents in coffee shops. Tiny teams are moving at a speed that would've sounded impossible five years ago. A quarter of Y Combinator's Winter 2025 cohort had codebases that were 95% AI-generated, founders who could have built everything themselves, but didn't need to.

This feels a lot like Apple's early App Store era. When Apple first launched the App Store, most people thought mobile apps would just be simple utilities like flashlights, calculators and weather apps. Nobody expected billion-dollar outcomes. Then Instagram happened, WhatsApp and Uber followed, and so on.

AI is doing something similar right now, except the cycle is moving even faster. Lovable, a Swedish vibe-coding startup, went from $1 million ARR to $100 million in annual recurring revenue (ARR) in under eight months. We're seeing startups from YC and a16z's Speedrun hit meaningful revenue numbers in months, not years. Small teams are reaching millions in ARR faster than traditional SaaS startups ever could.

The bottleneck is no longer building software. It is getting people to care. I believe this is the part most founders underestimate. Every week, thousands of new AI tools launch. Your feed is flooded with AI copilots, recruiters, SDRs, designers, assistants, note takers, browsers and video editors. Even though code became abundant, attention did not. When everyone can build, distribution becomes the moat.

I hear this constantly from the founders I mentor. Unless the product is genuinely novel, it's becoming brutally hard to be the 10th company in the same category.

Take AI notetakers as an example. When the category first emerged, there was plenty of room for experimentation. But over time, a handful of companies consolidated mindshare, integrations, partnerships, SEO authority and user trust.

Having spent over five years building in this space, I understand the years of iteration, integrations, workflows, enterprise capabilities, distribution loops and product depth that compound over time. From the outside, a new entrant might think, "I can build an AI notetaker in a weekend." And technically, they can. But building the product is only 5% of the challenge. The hard part is earning mindshare in a category where customers already associate the market with a few established players.

That same dynamic is now playing out across nearly every AI category, and what's even more interesting is that large incumbents aren't automatically winning either. You'd assume the big software companies with massive installed customer bases would dominate AI. But public SaaS markets tell a different story. Investors started questioning whether traditional SaaS products are defensible in a world of AI agents. The phrase "SaaSpocalypse" started circulating as many public software companies saw their valuations collapse. SaaS noted that software now trades at a discount to the S&P 500 for the first time in modern history, and the market started questioning whether seat-based software models would survive at all.

The fear is simple: Why pay for 10 dashboards if an AI agent can execute the workflow for you? Gartner predicts 40% of enterprise applications will be integrated with task-specific AI agents by the end of 2026, up from less than 5% in 2025. If agents replace the work, why keep paying per seat for the humans who used to do it?

Some of those concerns are legitimate. A lot of legacy software was built around humans manually operating interfaces. AI changes the interface layer entirely. But the market is overreacting when it says software is dead. Software isn't disappearing. The abstraction layer is changing.

The winners of the last era built systems of record. I believe the winners of the next era will build systems of action: agents that execute work and actually understand context rather than just retrieving it, software that runs continuously in the background without anyone clicking a button. But nobody has fully figured it out yet.

Yes, companies can now vibe code lightweight internal tools instead of buying SaaS products. A 2026 Retool survey found 35% of enterprises had already replaced at least one SaaS tool with an internal build. But who maintains those tools six months later? Who handles uptime, security, integrations, compliance and support? Most internal software eventually breaks down because maintenance is much harder than prototyping.

That's why software companies still matter. Building software is not hard anymore, but maintaining and scaling is. Consumer goods companies figured this out decades ago. Water is technically a commodity. Coffee is technically a commodity. Yet people still pay premiums for brands. Why? Taste, trust, identity, distribution and consistency.

I see software moving in the same direction. If code becomes commoditized, then everything around the code becomes more important. Brand, distribution, taste and community all begin to matter even more.

Success in this era won't necessarily mean having exclusive access to better AI models. I recommend understanding users deeply, moving faster than incumbents and building products people trust enough to adopt into daily workflows.

I don't think SaaS is dead. I think SaaS is evolving into agents. It reminds me of the transition from on-prem software to cloud software in the 2000s. Salesforce didn't invent CRM. They rebuilt it for a new platform shift. The incumbents had baggage while the new entrants didn't.

The next generation of great software companies will likely be built around AI from day one. But when anyone can build, standing out becomes the hard part. The next decade won't belong to the fastest builders. It'll belong to the companies that customers already rely on when it matters—the ones that earned trust before everyone else caught up.

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?

Read Entire Article