Merantix, BlueBridge, NEXT AI & Balderton: Moats in the Age of AI Separating Winners from Wannabes

Panelist Introductions & Business Models

• Solène (Baldur Capital) leads the discussion as an investor in European AI founders from Pre-seed to A-round.
• Bluebridge (represented by Carlos) defines itself as an "AI-native system integrator," positioning itself as a 3x faster and cheaper alternative to traditional integrators like Accenture or Deloitte.
• Mirantics (Nicole) operates a dual model: a service arm (Mirantics Momentum) building end-to-end AI solutions and a pre-seed investment arm (Mirantics Capital) based in Berlin and London.
• Next.ai (Moody) is a customer intelligence platform that unifies unstructured interaction data (sales calls, support tickets, reviews) to inform decision-making.

Definitions of AI Moats: Core Perspectives

• Bluebridge: A moat is defined by business value, flexibility, and modularity, allowing clients to swap underlying models as they evolve while maintaining high ROI through automation.
• Mirantics: "Human engineering" constitutes 70% of a moat, focusing on building sticky applications where users are "addicted" to the daily utility rather than just tech performance.
• Next.ai: Moats are built defensively through the process of solving hard customer problems, creating a "strong pipe" for data flow that accumulates sunk costs for customers if they switch, rather than owning the raw data itself.

Evolution of the Moat Landscape (Past 12–18 Months)

• Unstructured Data: AI now automates previously human-restricted steps involving voice, video, and text (e.g., contract reading), breaking the "human in the loop" dependency.
• Model Volatility: Leaders change frequently (e.g., Runway vs. Sora), forcing founders to build flexible pipelines that can swap models without breaking the product.
• Commoditization Fear: Founders worry about becoming mere features on top of large models, leading to a "liminal" state where market dynamics and customer access mechanisms (e.g., via glasses vs. browser) are undefined.
• Data Portability: Regulatory pressures for data portability may eventually lower switching costs, threatening proprietary data sources as a permanent moat.

Strategic Decisions & Operational Realities

• Engineering Layer as Moat: Next.ai emphasizes that "wrappers" are insufficient; substantial engineering is required to handle legacy scale (e.g., 1.5 million support tickets), which becomes defensible through complexity and proprietary processing logic.
• Systems of Record Debate:
  • Carlos: Argues against replacing legacy systems (SAP, Salesforce); instead, AI should replace the human tasks performed on top of these records to avoid creating redundant, expensive "layer cakes."
  • Consensus: Switching costs for legacy ERP systems are too high; successful AI strategies augment or automate workflows within existing records rather than duplicating them.
• Team Over Technology: Early-stage signals for moats include team adaptability, tenacity in finding a product wedge, and the ability to spot and react to problems rather than static technical advantages.
• Internal Culture: Next.ai founders admit they never hold internal meetings specifically about "defending a moat," preferring to focus on hustling, selling, and solving problems, trusting that defensibility emerges organically.

Signals for Investors (Mirantics Capital)

• Team Dynamics: The primary focus is on the team's ability to find a wedge, endure the discovery process, and execute with obsession (e.g., portfolio company Libra AI scaling ARR to $1M+ in six months after pivoting).
• Distribution & Brand: Despite AI's technical shift, distribution, brand, and first-mover advantage remain critical, particularly in high-barrier sectors like healthcare where "selling to the sector" is a significant hurdle.
• Sector Specifics: Specialized "deep tech" biotech applications currently offer stronger moats than generative AI wrappers due to the difficulty of translating wet-lab skills to the real world.

Q&A Insights & Forward-Looking Statements

• OpenAI's Moat: Moody acknowledges OpenAI's current lead (800M monthly users, brand recognition, vertical integration into hardware) but views it as a temporary snapshot, citing the "browser era" precedent where Google (not the first mover) eventually won.
• Human-in-the-Loop: Despite automation, strategic direction and the ability to distinguish "agile AI implementation" from traditional "SAP-style" rollouts remain fundamentally human tasks requiring intuition and experience.
• Regulation as Strategy: Building a moat solely on regulatory compliance (e.g., IP-cleared models) is deemed dangerous and unsustainable; businesses that bet entirely on owning their training IP have already lost their defensibility.
• Netflix Precedent: Netflix is highlighted as the only historical example of a company successfully reinventing its moat repeatedly (physical stores → mail → streaming → production → advertising → AI).

Advice for Founders

• Action over Frameworks: Founders should "rubber to the road," avoiding theoretical moat planning in favor of solving genuine, scalable customer problems and listening to user feedback.
• Target Substance: Investors and founders are encouraged to target "big, burning problems" in sectors like medical, engineering, and production where hype will fade but substance endures.
• Founder Vision: In small teams, the founder's vision is paramount; they must decide who to listen to (customers) and who to ignore, as clients will pull the product in conflicting directions.