Why Businesses Are Rejecting the AI They’ve Asked For ft Agency CEO Elias Torres

Core Philosophy on AI Adoption:

• Current market failure stems from an "expectation mismatch" where users demand 100% perfection from AI, contrasting sharply with the "narrow functionality, perfect accuracy" of pre-AI databases.
• Elias Torres argues that users should accept "broad functionality with imperfect accuracy," noting that a tool delivering 10-30% efficiency gains is valuable even if not flawless.
• He identifies "human friction" as the primary barrier to scaling customer experience, citing that employees often waste 20 minutes editing AI drafts to achieve perceived perfection, thereby negating the tool's speed benefits.

Agency (Company 3) Strategy & Mission:

• Mission: Build an "AI-led customer experience" platform capable of scaling interactions without requiring human intervention at every step, unlike the human-heavy models of the past.
• Product Focus: Unlike competitors focusing on front-end chat interfaces (e.g., email drafters, summarizers), Agency targets the "back end" and internal inefficiencies, aiming to automate entire workflows rather than just assist humans.
• Current Traction: Currently serves over 50 customers (including enterprise names like Klaviyo), with a strategic decision to prioritize making existing users "deliriously happy" over rapid new customer acquisition.
• Monetization Model: Torres explicitly rejects the immediate "load-on-browser and charge a dollar" mentality common in early SaaS, stating he seeks deep obsession and sticky usage before pursuing aggressive revenue targets.
• Financial Goal: Aims to build a $1 billion revenue company with fewer than 100 employees, shifting cost structures away from large GTM and Customer Success teams toward engineering.

Historical Context & Founder Trajectory:

• Early Life: Grew up in Nicaragua during a scarcity-driven "communist era," experiencing food insecurity and learning to ask for help, which shaped his work ethic and perspective on resource allocation.
• Career Path: First professional role at IBM (1999) building a "blue pages" chatbot; left due to bureaucracy (400,000 employees) to join David Cancel's startup Lookery (10 employees), where he lived and worked 24/7.
• Performable & HubSpot: Founded Performable, which was acquired by HubSpot in 2011; led the rebuild of HubSpot's core platform, scaling support from 20 to 5,000 customers and learning "hiring at scale."
• Drift: Co-founded Drift to solve "time to value" by automating lead capture and sales meetings; pivoted multiple times before finding product-market fit.
• Drift Lessons: Realized that scaling headcount (reaching ~800 employees) distracted from customer intimacy and product depth, leading to high churn and a realization that "hiring bodies" is a flawed scaling strategy for complex products.
• Return to Venture: Took a 7-month sabbatical in Brazil; upon ChatGPT's launch in late 2022, re-entered the startup ecosystem to build Agency, driven by a desire to solve the "impossible" problem of human-scaled customer experience.

Talent & Leadership Philosophy:

• Hiring Criteria: Prioritizes "hunger," "grit," and "intelligence" over credentials or pedigree; explicitly rejects candidates with extensive, perfect resumes, viewing them with suspicion ("sus").
• Impact: Credits hiring exceptional talent (e.g., Andrew Bilecki, Whitney) for HubSpot's success, noting that he often gave people chances they didn't initially have the credentials for, then modeled what "outlier" performance looks like.
• Growth Model: Encourages employees to take ownership of their paths, aiming to build a culture where people are shaped by the mission rather than just job descriptions.

Market Trends & Future Outlook:

• AI Capability Gap: Torres notes that while AI is theoretically capable, current Large Language Models (LLMs) struggle with consistency; requiring human prompting for every task renders them useless for true automation.
• "Agency" of Human Role: Optimistic that AI will liberate humans to focus on high-level "agency" (making decisions, choosing purpose, serving customers) rather than repetitive execution.
• Cost Concerns: Acknowledges the current LLM cost structure is unsustainable ("a shit show," "pyramid scheme") and is a significant operational challenge.
• Deprogramming the Market: Views the primary challenge as "deprogramming" businesses that rely on manual workflows; convincing them to trust AI to execute complex tasks (e.g., auto-emailing customers for renewals) without human oversight.