B2B SaaS Is Dead: How AI Founders Win in 2025

The rules of building a startup have fundamentally changed. Pablo Srugo, seven-year VC and founder since 2013, argues that GenAI didn't just add new tools — it invalidated the entire late-stage B2B SaaS playbook and opened a new garden of low-hanging fruit for founders willing to think differently.

Key Facts

• Era dividing line: ChatGPT launch (late 2022) marks the shift from incremental SaaS to undeniable AI value (Pablo Srugo)
• Triple threat competition: Founders now compete against incumbents, foundation models (Anthropic, ChatGPT, Gemini), AND fast new entrants simultaneously (Pablo Srugo)
• Time-in-market moat: No longer valid — a new entrant today can match a two-year-old AI product out of the box due to model improvements (Pablo Srugo)
• Service-as-Software example: TENEX grew from $0 to $40M ARR in one year by becoming an AI-first managed security provider instead of selling AI to MSPs (Pablo Srugo)
• Core pricing shift: Per-seat SaaS is no longer the default — usage, transactional, per-output, and outcome-based models are all live (Pablo Srugo)

Why the Late-Stage B2B SaaS Playbook Stopped Working

By 2018–2022, most SaaS opportunities were incremental — niche verticalizations of tools that already existed. Value props were marginal time-saves, not 10x leaps. Investors had to constantly ask whether a product was truly must-have or just a nice-to-have with a small, addressable niche.

Pablo Srugo traces the arc clearly: Salesforce brought CRM to the cloud. HubSpot owned inbound. ServiceTitan and Jobber took large verticals like home services. By the late 2010s, the remaining opportunities were things like a CRM built specifically for recruiters — real, but marginal.

The value unlocks in this era were mostly time-saves on tasks that were already a small part of someone's job. Srugo describes it plainly: 'What used to take you four hours a week now takes you two hours a week. So I've saved you two hours a week.' That's not a company-defining insight — it's a feature.

This mattered enormously for founders pitching at seed. The number-one filter Srugo applied was whether the value prop was truly undeniable. Most weren't. Incremental products are painful to sell — and even harder to scale on a venture timeline.

"The low hanging fruit is gone. And you start having to find smaller, and smaller, and more, and more niche opportunities." — Pablo Srugo

"Pushing a nice-to-have product is really pushing a boulder up a hill and it's a painful, painful exercise." — Pablo Srugo

• Salesforce → HubSpot → ServiceTitan: each wave picked smaller fruit
• Late-era SaaS value props were mostly marginal workflow improvements
• Nice-to-have products stall growth and lead to over-raising
• 'Pushing a nice-to-have product is really pushing a boulder up a hill' — Pablo Srugo

GenAI Unlocks Undeniable Value — But Multiplies Competition

Post-ChatGPT, delivering genuinely transformative value is now the baseline, not the differentiator. The hard part has flipped: generating undeniable value is easier than ever, but competition now comes from three directions at once — incumbents, foundation models, and fast-moving new entrants.

The clearest example Srugo uses is voice AI. Workflows that required a $20/hour human — logistics call centers, 311 non-emergency dispatch — can now run at roughly $2/hour with AI. 'In terms of value unlock, that's completely undeniable,' he says. That's a 10x cost reduction, not a marginal efficiency gain.

But that same technological leap has made competition brutal. Incumbents can add AI to their existing products. Foundation models like Anthropic, ChatGPT, and Gemini expand their native capabilities weekly. And new entrants can match a two-year-old product on day one, because the underlying models have improved dramatically in the interim.

Srugo is direct about the tradeoff: 'As a founder I would much rather be in a situation where I'm delivering insane value and have to compete, than a situation where my product might be nice to have and I don't have any competitors.' The competitive intensity is real — but so is the tailwind.

"There are so many places where what used to cost $20 an hour, because it was a human. Now is going to cost $2 an hour, because it's an AI. In terms of value unlock, that's completely undeniable." — Pablo Srugo

"You always have to worry about: is the thing that you're doing differentiated enough, integrated enough, verticalized enough — such that it's unlikely that the massive foundational models are just going to do what I do?" — Pablo Srugo

Time-in-Market Is Dead — Cycle Speed Is the Only Moat

A two-year head start in AI no longer guarantees a defensible lead. Model improvements mean new entrants can match established products out of the box. The only durable advantage is cycle speed: how fast you ship, test, and iterate — combined with a deeper understanding of your ICP than any competitor.

In the pre-GenAI era, being first to market with a reasonably featured product created a real moat. A competitor starting two years later had a worse product, weaker brand, and less distribution. That math no longer holds. If you built a voice AI product two years ago, you likely had to construct scaffolding around early models just to make them production-ready. A new entrant today gets that capability off the shelf.

Srugo uses a Ben Thompson analogy to explain why staying ahead matters continuously: Thompson's subscribers don't pay for any single article — they pay for a constant stream. 'If your product is always one or two months ahead than competitors, it will always continue to be an edge,' Srugo says. Customers signing annual contracts are buying the product roadmap, not just the current feature set.

The practical implication: cycle speed and ICP knowledge are now your primary competitive weapons. Founders who understand their customers deeply will build the right thing more often, waste fewer cycles, and maintain that one-step lead that compounds over time.

"Your time in market moat frankly doesn't exist. So now you've got to compete with incumbents, foundational models and new entrants." — Pablo Srugo

"One of the ways to stay on top is to just be that much faster than your competitors — and that has two pieces. One is literally the cycle speed. The second piece is how well do you understand your ICP." — Pablo Srugo

Service-as-Software: The Massive Opportunity Most Founders Are Missing

AI enables founders to build scalable services companies — not just SaaS tools. Instead of selling AI to service firms, you become the service firm, AI-first. TENEX did this in cybersecurity, growing to $40M ARR in one year by replacing managed security providers rather than selling to them.

The traditional VC logic held that services businesses were unfundable: low margins, hard to scale, quality degrades as headcount grows. Software solved that by being infinitely replicable. AI has now extended that logic to services themselves. You can deliver a professional service — security monitoring, fractional CFO, accounting — at software-like scale and improving margins.

TENEX is Srugo's anchor example. Rather than selling AI productivity tools to managed security providers (MSPs), the founder became an MSP — AI-first. The pitch to customers: 'You're paying half a million, a million dollars a year today for an outsourced MSP. I'm going to do the job way better than them, have more coverage, resolve tickets faster, and do it twenty to thirty percent cheaper.' Customers can make that decision instantly because you're solving the whole problem.

Srugo sees this pattern repeating across verticals. An AI-first fractional CFO that closes the month in one day instead of fifteen, at $4,000/month instead of $5,000, is a straightforward displacement. The demand already exists. You're not creating a new category — you're entering a proven market with a structurally superior cost base.

"What AI is doing is totally changing the game — it's allowing founders to all of a sudden deliver services. Create services companies that are AI first." — Pablo Srugo

"The world has really changed and few are the founders who are really thinking from first principles, and understanding what has now opened up. If you were one of the first, it's a massive edge." — Pablo Srugo

• TENEX: $0 to $40M ARR in one year as an AI-first managed security provider
• Target any services business delivering a digital end product: legal, finance, accounting, security
• Compete on price and quality, not on selling productivity software to incumbents
• Existing market demand means no category creation required — faster go-to-market

New Revenue Models: Beyond Per-Seat SaaS Pricing

Per-seat monthly pricing is no longer the default. AI-native startups are experimenting with usage-based, per-output, transactional, and outcome-based models. The right model depends on how value is actually delivered — and founders who default to old pricing structures leave significant revenue on the table.

In the final era of B2B SaaS, pricing was simple: users times monthly fee, with annual discounts. Srugo acknowledges this clarity was useful, but it's no longer the only or best option for AI products. The models now in play are diverse: per-agent, per-conversation, per-minute, usage/token-based, or straight outcome pricing.

For voice AI specifically, Srugo lays out the full menu: charge by number of AI agents deployed, by seats in the call center using AI as a pre-filter, by completed conversation, or by minute of AI interaction. Each model aligns differently with customer value and usage patterns.

The broader principle: pricing should reflect how the AI actually delivers value. If a customer gets ROI per resolved ticket, charge per ticket. If value scales with usage volume, token pricing makes sense. Founders who default to per-seat out of habit are pricing against their own value proposition.

"Today, everything is up for grabs. There are people that are still charging on a per user per month basis. There are people talking on a usage basis like token basis. There are people charging transactional. There are people charging on the delivery of the ultimate output." — Pablo Srugo

Pre-GenAI B2B SaaS vs. Post-GenAI AI Startups