NGE · Investment Letter · Issue 33 · June 2026

Second-Mover
AI.

AI summarization, AI-powered insights, and smart automation features are table stakes by 2026. They improve retention at the margin. They do not change the structural question of whether your product survives the commoditization of its core workflow. The companies that win the next decade in legacy industries won't be the ones who shipped the first chatbot — they'll be the ones who own data nobody else can access.

Not investment advice. Not a recommendation to buy or sell. Research and long-horizon thinking only. Consult a qualified financial advisor before making any investment decision. Figures cited are sourced from Andreessen Horowitz, Bessemer Venture Partners, HarbourVest, and Gartner, current as of writing.

The False Comfort of "AI Features"

Adding a chatbot
is a delay tactic, not a strategy.

Every SaaS company is adding AI features right now — AI-powered insights, smart summaries, automated workflows, predictive analytics. By 2026, these are table stakes, not differentiators. A buyer evaluating two project management tools is not choosing based on which one has AI summarization; they are choosing on price, integrations, and switching cost. Adding AI summarization to a note-taking tool does not create a data moat. Building a smart assistant on top of a CRM does not create compliance infrastructure. The most dangerous response to AI disruption, according to current private equity analysis, is precisely this — adding AI features to a product that has none of the three structural moats that actually survive: proprietary data, regulated-industry compliance infrastructure, and deep systems-of-record integration.

The scale of the underlying disruption is real and measurable. Retool's 2026 Build vs. Buy Report, surveying 817 enterprise software builders, found that 35% have already replaced at least one SaaS tool with a custom-built alternative, and 78% plan to build more in 2026 — workflow automation, internal admin tools, and business intelligence are the categories under the most pressure. Gartner's strategic predictions go further: by 2028, 90% of B2B buying will be AI-agent-intermediated, pushing more than $15 trillion of B2B spend through AI agent exchanges rather than human-navigated software interfaces.

"Software ate delivery. AI eats the task itself. The SaaS wave was a distribution story. This wave is a replacement story." — Attainment Labs, February 2026

35%
Enterprises that already replaced a SaaS tool with custom AI-built alternative
$15T
B2B spend projected through AI-agent exchanges by 2028
78%
Enterprises planning to build more AI replacements in 2026
The Three Moats That Survive

Data, compliance, and integration —
not the model itself.

Private equity operating partners are converging on the same framework: as AI capabilities become commoditized and increasingly accessible to any well-funded competitor, defensibility shifts entirely to what surrounds the model rather than the model itself. A data moat is created when a company owns and continually enriches proprietary data that competitors cannot easily access or replicate. The metaphor used by Andreessen Horowitz partner Andrew Rampell is precise: if a restaurant sources exclusive, high-quality vegetables, it can charge a premium for the final meal because competitors lack the raw materials — in software, this translates to controlling obscure or non-public data sources that large research labs cannot easily procure.

Legacy "Technophobic" Verticals

The industries everyone assumed would resist AI are adopting it fastest

Restaurants, contractors, freight, dentists — operational criticality protects against disruption

Bessemer Venture Partners' conviction has only strengthened: vertical AI has the potential to eclipse even the most successful legacy vertical SaaS markets, with adoption accelerating particularly in workflows long considered manual, service-heavy, and resistant to technology. The pattern that emerges across construction, wellness, freight, and dentistry is consistent — these industries do not cut software spend dramatically in downturns, and loops reinforce market share: in construction, subcontractors adopt whatever tool general contractors use; in wellness, consumers book through whichever marketplace already has the network. Operational criticality plus embedded fintech — payment processing fees adding 2-3x revenue uplift per customer compared to pure SaaS — has become a key defensive layer specifically because payments volume is not affected by AI-driven seat compression the way pure software subscriptions are.

90 deals
Companies scored on data moat + compliance + integration gravity (max score 9)
2.4x+
Minimum revenue multiple for companies scoring 7+ on the moat framework
$7T
US embedded finance transaction volume forecast by 2026
Where the Capital Is Flowing

Specific examples,
not just a framework.

⚖️ Specialized Legal Data

Vlex, cited directly by a16z as a model example, aggregates specialized legal datasets that large general-purpose research labs cannot easily procure or replicate, then wraps that proprietary data in an inference layer that delivers outputs no general model can match.

🩺 Clinical Evidence Data

OpenEvidence applies the same data-moat logic to medicine — proprietary, curated clinical datasets that become more valuable and more defensible the longer the company operates and accumulates additional verified data points.

🏪 Retail Transaction Data

A retailer's point-of-sale transaction data, flowing in real time into a governed data lakehouse, becomes the proprietary fuel for demand forecasting and dynamic pricing agents — value that strengthens specifically because the data is governed and actionable, not generic.

🤝 The Collaboration Layer

a16z's Alex Immerman frames the next network-effect frontier as multi-party collaboration: when value increases from multi-human and multi-agent collaboration across buyers, sellers, tenants, and vendors, switching costs rise — the collaboration layer itself becomes the moat.

Why "Second-Mover" Is the Right Framing
The companies that win this cycle are explicitly not the ones who shipped the first AI feature. Composable, API-first, cloud-native architectures establish the durable competitive moat, according to Gartner's strategic predictions — and IDC's 2025 analysis found that companies with modular, composable architectures reduce total cost of ownership by up to 37%. Being second to market with a deliberately data-moated, compliance-anchored product beats being first with a feature any competitor can replicate within a quarter.
The Repricing Already Happened Once
The February 2026 software-sector selloff is a direct preview of this risk materializing in public markets. Anthropic's February 2026 release of autonomous legal and business workflow capabilities made the AI-replaces-software thesis a measurable reality rather than a theoretical debate, and share prices of legal-tech, market-data, business-intelligence, and horizontal SaaS tools were repriced sharply downward within days. Any company without one of the three structural moats — proprietary data, compliance infrastructure, or genuine integration gravity — should be assumed vulnerable to a similar repricing event, not treated as a hypothetical risk.
The Honest Read

The data-moat framework is becoming consensus among sophisticated investors precisely because it is intuitive and well-articulated, which raises the question of whether it is already priced into the winners. Companies scoring 7+ on a16z and PE operating partners' moat frameworks already trade at meaningful premiums — 2.4x revenue at minimum. The genuine alpha in this thesis likely sits in correctly identifying moated companies before that recognition becomes fully consensus, not in confirming what venture capital firms are already broadcasting in their public 2026 outlook reports.

The timeline for building a defensible position is explicitly narrowing, according to the same analysts making this case, which cuts both ways for investors. If the window for incumbents to build genuine data moats before well-capitalized AI-native competitors arrive is closing, that argues for urgency in identifying winners now — but it also means a meaningful share of currently-promising vertical AI companies will fail to establish defensible moats in time, and distinguishing the two in advance remains genuinely difficult even with this framework in hand.

The NGE View

The verdict.

What We Believe
AI features without a data moat are not a strategy — they are a confirmed delay tactic, evidenced by the February 2026 repricing event. Any software company whose primary AI value-add is summarization, automation, or insight-generation without proprietary data underneath it should be evaluated as structurally exposed, not defended.
Legacy, "technophobic" verticals are the most underrated opportunity in this entire AI cycle. Restaurants, contractors, freight, and dentistry were never actually resistant to technology — traditional SaaS simply failed to solve their multi-modal, language-heavy, high-value tasks, and vertical AI is now meeting them precisely where they were underserved.
Embedded fintech is an underappreciated defensive layer worth weighting specifically. Payment processing revenue uplift of 2-3x per customer, combined with payments volume being immune to AI-driven seat compression, makes embedded-fintech-plus-data-moat companies meaningfully more durable than data-moat-alone companies.
Treat the three-moat framework as a screening tool, not a guarantee — the real work is verifying data proprietary claims directly. Many companies will claim a "data moat" that is, on inspection, a static transaction record rather than an enriching, compounding dataset — the distinction between the two is the entire thesis.

The companies most exposed in this cycle are not the obvious legacy laggards — they are the companies that moved fastest to bolt AI features onto products with no underlying defensibility, mistaking visible activity for genuine strategic positioning. The companies best positioned are the patient second movers: the ones quietly accumulating proprietary data in unglamorous, overlooked verticals, building compliance infrastructure in regulated industries with long switching cycles, and embedding themselves so deeply into operational workflows that removal becomes genuinely catastrophic for the customer. Being early to ship an AI feature was never the moat. Owning what the AI needs to actually work always was.

NGE · A Futuristic Investment Letter

Long-horizon thinking on capital, technology, and the forces shaping the next decade of wealth creation. Written from first principles. Not consensus. Not noise.

— Pawan Bhatia · NextGen Economics · Bangalore, India