Interview, Fireside Chat
"We Don't Trust Agents" - What This CTO Knows That You Don't
- Customer Base & Market Position: Merge serves major enterprise and AI customers including OpenAI, Perplexity, Netflix, Uber, Mistral, Dropbox, and Freshworks, powering their core search functionality and product connectivity.
- Scale Metrics: The platform supports over 25,000 self-serve organizations and more than 400 enterprise customers across financial services, SaaS, and AI sectors.
- Service Level Agreements (SLA): The company maintains four to six nines of uptime criticality, acknowledging that any downtime could result in immediate loss of enterprise customers.
- Product Suite Evolution: Merge recently expanded from a single Unified API into a three-product suite:
- Merge Unified: Normalizes data ingestion for enterprise search and RAG (Retrieval-Augmented Generation).
- Merge Agent Handler: A single Model Context Protocol (MCP) server connecting agents to hundreds of integrations for internal and external use cases.
- Merge Gateway: Infrastructure for routing between hundreds of AI models with security layers, smart routing policies, and cost-optimization dashboards.
- Strategic Pivot: A year and a half ago, the leadership paused a major deal consuming 90% of resources to pivot the business toward AI, utilizing the downtime to build the Merge Agent Handler with a team of just one engineer and the founders.
- Internal AI Adoption: The entire company, including non-technical teams (accounting, recruiting, finance), aggressively uses AI coding tools like Cloud Code and Windsurf to accelerate development; 2023 revenue accelerated significantly without a corresponding headcount increase.
- Hiring Philosophy: Merge prioritizes "missionaries" over "mercenaries," seeking candidates with a demonstrated desire to use AI and build things from zero to one, even if current employers restrict AI usage.
- Security Threat Landscape: The industry faces surging AI-driven supply chain attacks where agents inject vulnerabilities into open-source packages at a volume humans cannot manually review; one single breach could be fatal for a company.
- Bot Attack Volume: Merge recently encountered an automated bot attack exceeding 1,000 signups in a single hour, scanning all backend endpoints from global IPs, highlighting the difficulty of blocking AI-driven reconnaissance.
- AI Security Vulnerabilities: Security experts note that AI agents now "speak English perfectly, write code perfectly, and have unlimited manpower," making them indistinguishable from human attackers and capable of mass-producing exploits.
- Data Governance Strategy: Merge enforces hard guardrails to prevent non-deterministic agents from leaking sensitive data (e.g., Social Security numbers) to external tools, moving beyond trust-based models to rule-based blocking.
- Sales Cycle Shifts: Large financial services clients have longer sales cycles, whereas large AI companies purchase rapidly due to fierce competition and immediate product readiness; however, AI buyers often lack technical expertise on protocols like MCP.
- Infrastructure Security: The company invests heavily in internal security, prioritizing background checks and reference calls because internal human compromise is viewed as a greater threat than external scanning.
- SaaS Disruption Trend: As AI coding capabilities improve, the cost for enterprises to build software in-house decreases, creating pressure on traditional SaaS platforms to offer higher levels of automation and "headless" integration capabilities.
- Headless API Adoption: Companies are increasingly adopting "headless" operations where agents manage software via API/CLI (e.g., Salesforce) rather than UI, reducing the barrier to entry for building and managing tools.
- Valuation Sentiment: The founders express skepticism regarding the 300x to 1000x revenue multiples seen in the current AI private market, predicting that many overvalued companies will face layoffs or funding gaps despite initial success.
- Misconception on Model Building: A prevailing industry trend of companies building and training custom models is deemed an over-engineering mistake; the consensus is that generic models combined with strong product execution are more effective.
- Future Roadmap: The company anticipates 2024-2025 will be the year enterprises move from "allowing ChatGPT usage" to actively deploying governed, production-grade AI agents for internal automation.
- Revenue Focus: Primary business metrics are cash burn, runway, gross margins, and cash burn multiples, with a strict mandate to avoid second-place market positions and immediately act on negative feedback.
- Salesforce Partnership: Merge views Salesforce's shift toward headless operations as a validation of their approach, citing Salesforce's 30-year history of API-first growth and Benioff's "beginner's mind" philosophy.
- Customer Onboarding Evolution: Early-stage startups were hesitant to use Merge due to product infancy, but the company successfully climbed the "logo ladder" to win enterprise trust through rigorous reliability and enterprise-ready features.