No PriorsNo Priors Ep. 94 | With CEO and Founder of Agency Elias Torres
CHAPTERS
- 0:00 – 1:26
Elias Torres’ origin story: Nicaragua to Boston, IBM to entrepreneurship
Sarah introduces Elias Torres and his track record (HubSpot engineering leader, Drift exit, now building Agency). Elias shares his immigrant background, early career at IBM, and the drive for freedom and impact that pushed him toward startups.
- •First-generation immigrant from Nicaragua; early struggles with English and resources
- •Spent ~10 years at IBM before choosing entrepreneurship
- •Motivation: autonomy, outsized impact, and building rather than operating inside big companies
- •Built career largely in Boston rather than Silicon Valley
- 1:26 – 2:40
Math competitions, problem-solving mindset, and adapting in the U.S.
They discuss Elias’ early math competition experiences in Nicaragua and later in Florida. Elias frames math as foundational training for structured thinking, while also describing the challenges of language barriers in school.
- •Represented his school nationally in Nicaragua; placed third in the country
- •Joined Mu Alpha Theta in a low-income Florida public school despite not speaking English
- •Math as transferable cognitive training for engineering and entrepreneurship
- •Early example of thriving through constraints (word problems vs. pure math)
- 2:40 – 4:22
HubSpot’s IPO journey and the “naivete advantage”
Sarah asks why Elias left a high-scale leadership role at HubSpot to start Drift. Elias describes learning what “going public” meant only by doing it, and how that experience created confidence—plus overconfidence—about repeating the outcome from scratch.
- •Performable acquisition led Elias into HubSpot at a pivotal growth stage
- •Goal to IPO felt abstract; learned by executing with product/engineering autonomy
- •Revenue growth narrative: ~$30M to ~$130M and IPO success
- •Naivete as a catalyst: believing zero-to-$100M would be straightforward
- 4:22 – 6:17
Selling Drift: why a $1B+ outcome felt like failure
Elias explains the emotional whiplash of selling Drift for over a billion dollars in cash. Compared to HubSpot’s continued growth, the goalposts shifted—so selling felt like losing the bigger dream of building a $30B public company, compounded by exhaustion.
- •Personal gratitude vs. ambition-driven disappointment can coexist
- •Goalpost shift after seeing HubSpot scale from $1B to far larger
- •Exit moment was anticlimactic; felt “nothing” at signing
- •Founder burnout and the feeling of being ‘done’ after nonstop work
- 6:17 – 9:36
ChatGPT as the catalyst: curiosity, urgency, and consulting with OpenAI
After a rare vacation, Elias encounters ChatGPT and feels he’s missed the next major wave. He connects with OpenAI, discovers they’re overwhelmed by customer implementation needs, and jumps in—starting with basic support and quickly moving to major client deployments.
- •October/November 2022: realization that LLMs change what’s possible
- •Initial gap: knew transformers/BERT but not practical LLM application via APIs
- •OpenAI’s scaling pain: many customers, limited internal implementation bandwidth
- •Hands-on learning: tickets → enterprise contracts (NBA, Ticketmaster, Red Bull, Klaviyo)
- 9:36 – 11:40
Why customer success became the wedge for Agency
Sarah probes how Elias landed on customer success as the focus. Elias traces it to work with Klaviyo’s CEO on scaling post-sales outcomes across hundreds of thousands of customers, and realizing LLM-driven systems could deliver benchmarking, insights, and proactive guidance at scale.
- •Customer success as an unsolved scaling problem in B2B SaaS
- •Klaviyo project: scaling CS for massive customer counts
- •LLMs enable detailed, personalized insight generation humans struggle to do at scale
- •Agency’s thesis: apply learnings across prior companies to post-sales execution
- 11:40 – 15:48
Customer experience in 5 years: put the customer in charge (at scale)
Elias argues “CS” is a misleading category label; the real objective is customer-centricity with high-touch quality at massive scale. He uses analogies (barber, local car storage) to describe flexible, relationship-driven service—then asks why enterprise software can’t feel like texting a trusted human.
- •Reframing: not ‘CS software,’ but ‘serve customers like they’re the only one’
- •Founder pain point: scaling from dozens to thousands of customers breaks intimacy
- •Analogies for ideal UX: direct text, flexibility, customer-led relationship
- •Agency goal: use AI to maintain high-touch service for hundreds of thousands of customers
- 15:48 – 18:31
Building a lean, AI-native company: 100 people, $1B revenue
Sarah asks how a modern company avoids scaling headcount to 800+ while targeting massive revenue. Elias emphasizes first-principles thinking, choosing big enterprise problems, and rejecting ‘do things that don’t scale’ in favor of designing scalable systems from day one—enabled by AI and experience.
- •North Star: $1B revenue with ~100 employees
- •First-principles approach (SpaceX/Elon analogy) to cost and structure
- •Two paths to $1B: massive volume/low CAC vs. high-price enterprise value
- •AI changes execution speed; build only scalable processes from the start
- 18:31 – 20:32
Raising the bar while shrinking the team: contractor-first hiring
Elias outlines a stricter hiring philosophy shaped by repeated mistakes and pattern-matching across executives he’s hired before. Agency requires candidates to prove they can deliver world-class output via short contractor trials, with clear roles and high accountability—especially for engineering.
- •Common founder error: over-indexing on pitches and pedigree
- •Experience hiring across many C-level roles drives increased skepticism
- •Policy: no one joins until they’ve delivered as a contractor in 1–2 weeks
- •Operating model: small team, explicit ownership, very high performance bar
- 20:32 – 23:03
Hardest parts of building Agency: product depth, trust, and change management
Elias argues LLMs solve only a small slice; the real work is building a product that fits workflows and shepherding organizations through adoption. Trust is the bottleneck—people may accept summaries, but they scrutinize AI-generated customer communications that carry real revenue risk.
- •LLMs are not the product; product design and workflows remain the core challenge
- •Enterprise adoption requires gradual transition, not ‘rip and replace’
- •Trust gap: AI can summarize, but sending emails to $1M customers is different
- •Change management and customer-by-customer learning determine success
- 23:03 – 26:25
‘Software enslaves us’: why CRM/data-entry workflows are broken
Sarah challenges Elias’ anti-software stance, and he responds that most software forces humans into repetitive data entry and bureaucracy. He critiques CRMs like Salesforce as costly, complex systems that create layers of administration instead of helping teams engage customers more effectively.
- •Provocation: workers are already ‘enslaved’ by bad workflow software
- •CRM critique: requires configuration, specialized admins, and constant data input
- •Organizational tax: people managing people managing software rather than customers
- •Positive model: software should be like Uber—simple intent → outcome
- 26:25 – 28:14
Next-gen software: proactive, invisible, and personalized—built from first principles
Elias describes the desired future: software that reads context (calendar, relationships, priorities) and takes action proactively, asking for verification rather than instructions. He predicts infrastructure providers will remain, but ‘database wrappers’ will die unless they reinvent around agentic, outcome-driven experiences.
- •Vision: software that does work automatically (outreach, scheduling, prioritization)
- •User value = ‘feel wealthy’: texting intent and having it completed end-to-end
- •Incumbents: infrastructure persists, but traditional UI/table/workflow layers must adapt or fade
- •Call to builders: stop making AI-skinned CRMs; rebuild software to be invisible and outcome-first
- 28:14 – 31:18
Closing reflections: let AI cover weaknesses; humans focus on relationships
Sarah relates the vision to her own workflow challenges and hopes software adapts to humans rather than the reverse. Elias argues people should lean into their strengths while AI handles organization and preparation—ending with a reminder that relationships and trust are what matter most.
- •Software should accommodate human variance (disorganization, context switching)
- •Personal development heuristic: don’t spend years fixing weakest bucket when tools can help
- •AI as a co-pilot for prep, context, and organization; humans provide judgment and rapport
- •Wrap-up: relationship-building remains the enduring differentiator
