Silicon Valley Girl$1.5B AI Founder: This Is Your Golden Age to Build With AI
CHAPTERS
Meet Decagon: AI agents handling real customer conversations at scale
Marina introduces Jesse Zhang and Decagon, a fast-growing AI company building conversational agents for large brands. Jesse explains what “automating with AI agents” means in practice: end-to-end customer conversations that can look up user context and take actions like booking or upgrading.
- •Decagon focuses on customer-facing conversations (phone/chat), not generic workflow automation
- •Agents retrieve customer data (history, loyalty tier) and execute actions (bookings, changes)
- •Example use cases: hotels, travel, consumer apps with high inquiry volume
- •Framing AI as operational leverage rather than a simple chatbot upgrade
Is AI taking jobs? Three adoption modes companies fall into
Jesse breaks down how companies deploy AI agents depending on their goals. Some use AI to scale without proportional hiring, others prioritize customer experience, and a third group focuses on cost reduction—sometimes cutting outsourced agencies.
- •Three buckets: growth amplification, quality/experience upgrades, cost-saving efficiency
- •In growth mode, AI reduces the need to scale support headcount linearly
- •In experience mode, AI improves speed and satisfaction rather than reducing staff
- •In cost mode, outsourcing agencies are most likely to be downsized
- •Roughly “one-third / one-third / one-third” distribution across these motivations
What happens to displaced work: from tier-1 support to higher-value roles
Rather than eliminating all work, Jesse argues AI shifts human effort up the complexity ladder. Agencies and support teams move from repetitive tier-1 requests to complex tier-2/3 cases and new tasks like data collection and AI review.
- •Routine inquiries (password resets, basic bookings) are easiest to automate
- •Humans shift to complex interactions and relationship-based work
- •New work emerges: labeling/collecting data, QA, reviewing and improving AI outputs
- •Historical pattern: big tech shifts change job nature more than they erase all work
Sponsor segment: HighLevel as an all-in-one automation platform
Marina shares an overview of HighLevel as a unified system for running a small business without stitching together many tools. The pitch emphasizes built-in AI “employees” for calls, messaging, content, reviews, and workflows.
- •Consolidates website/funnels, CRM, scheduling, payments, courses, and inbox
- •AI features: voice agent, conversation replies, review follow-ups, content generation
- •Positioned as simple setup and cost-effective starting at $97/month
- •Offer: extended trial via Marina’s link
The future of entry-level work: fewer openings, different job shapes
Marina asks about tightening entry-level markets; Jesse responds that demand won’t vanish but roles will evolve around AI-enabled productivity. He uses software engineering as an analogy: output demand is “uncapped,” while AI changes how work gets done.
- •Entry-level competition rising while roles shift due to AI leverage
- •Jobs won’t simply disappear; tasks and expectations change
- •Engineers heavily use AI already; productivity impact is real but hard to measure
- •Core idea: AI increases throughput, which can expand what organizations attempt to build
Tools people actually use: prototyping, coding assistance, and research
Asked for top tools, Jesse shares what his team uses rather than prescribing a universal stack. He highlights AI coding environments, rapid prototyping tools, and everyday productivity aids like ChatGPT and note-takers—more “amplify the worker” than “everyone builds agents.”
- •Engineering tools mentioned: Cursor, Cloud Code (coding with AI assistance)
- •Prototype/build tools: Lovable for quick iteration
- •Non-engineering: ChatGPT for research and go-to-market enablement
- •Most employees use AI to amplify work rather than building custom agents from scratch
Why Decagon starts with enterprises: scale, iteration, and product maturity
Jesse explains Decagon’s enterprise focus: large companies have the volume to justify investment and provide tight feedback loops. As the product matures, the long-term path is to “productize” for smaller customers who can’t support bespoke iteration.
- •Enterprises have high inquiry volume and clearer ROI for automation
- •Early-stage AI agents require rapid iteration and close collaboration
- •Working with large clients helps refine capabilities and reliability
- •SMB expansion typically requires a more standardized, productized offering
How Decagon uses AI internally: research and preparation compress dramatically
Marina asks what processes Decagon has automated inside its own team. Jesse points to research: AI tools now compress hours of Googling, note-taking, and context building into minutes with citations that can be validated.
- •Internal automation example: market/company research and customer context gathering
- •AI deep-research agents accelerate preparation for sales and strategy
- •Citations/sources help validate outputs and reduce hallucination risk
- •Outcome: faster learning loops and better customer empathy/context
Can non-technical founders build in AI? Technical helps, but it’s a golden time
Marina presses on whether founders must code to build a billion-dollar AI company. Jesse says technical skill accelerates decision-making and intuition, but AI tools expand what semi-technical and non-technical builders can accomplish—making now a uniquely favorable moment.
- •You don’t have to be technical, but it materially improves speed and judgment
- •AI tooling broadens the builder audience beyond traditional engineers
- •Learning technical basics is encouraged if you have interest/time
- •Thesis: “golden time” for non-technical people to enter and build
College grads: work first or start now? A decision framework
Jesse offers a pragmatic lens: start a company if you have conviction and energy, but expect it to be tough without intuition. If you work first, he recommends joining a startup that is post–product-market fit to learn what “good” looks like at scale.
- •Start now if you feel ready and can handle steep learning curves
- •Working first can build intuition about customers, teams, and execution
- •If choosing a startup job, prefer post-PMF to learn scalable product building
- •Post-PMF teaches customer dynamics, operating rhythms, and reliability at scale
Jobs most at risk: “straight output” roles and the shift to AI-guided work
Asked which jobs may go extinct, Jesse points to roles defined primarily by producing output that models can now generate—like writing marketing copy. He argues roles won’t vanish entirely; they evolve into supervising, guiding, and shaping AI output toward business goals.
- •Highest risk: jobs where the value is primarily raw content/output generation
- •Example: pure copywriting becomes “AI-operated” marketing content production
- •Support roles shift from answering basics to managing AI + handling complex cases
- •General pattern: humans guide and steer technology rather than replicate it manually
New career paths: conversation/AI architects and enablement programs
Jesse describes emerging roles at customers like “conversation architect” or “AI architect,” responsible for designing agent behavior. Decagon trains these users via “Decagon University,” turning CX managers and chatbot/knowledge-base owners into AI-era operators.
- •New role archetype: design how AI behaves in real customer interactions
- •Skill emphasis: reasoning, clarity, and instruction-writing in natural language
- •Many roles evolve from CX managers, knowledge-base owners, and chatbot leads
- •Upskilling and enablement become core: Decagon University as a transition pathway
What employers want now: analytical thinking + crisp communication
Marina asks what Jesse looks for when hiring in the AI era, especially for people transitioning from support to AI management. Jesse emphasizes the ability to analyze failures step-by-step and communicate unambiguous instructions—because “teaching” AI is fundamentally a communication task.
- •Analytical trait: break down interactions, diagnose why the AI responded that way
- •Tooling helps trace reasoning, knowledge used, and steps taken in conversations
- •Communication trait: write clear, non-contradictory instructions in plain English
- •These skills enable people to improve systems continuously, not just execute tasks
Finding the right startup idea: customer discovery, revenue signals, and B2B vs B2C
Jesse explains Decagon’s origin: intensive customer conversations to find a quantifiable, high-ROI pain point. He discusses why they chose large B2B (more predictable, easier to reason about) and outlines how founders should validate with willingness-to-pay and decision-process details.
- •Core method: stay extremely close to customers to find “ground truth” needs
- •Prefer problems with measurable ROI and clear impact metrics
- •B2B chosen deliberately: more predictable than B2C, easier to validate value
- •Validation steps: get pricing anchors, identify budget owner, map decision-making process
Advice for future AI founders: don’t copy paths—gather signal and learn sales
Jesse closes with founder guidance: avoid over-indexing on other founders’ playbooks; tailor your approach to your strengths and circumstances. He stresses that early-stage building is mostly signal gathering (often via sales-like discovery) and that founders should delay building until they have strong validation.
- •Don’t blindly replicate other founders—optimize for your own strengths
- •Early stage is about maximum signal: customer conversations and concrete commitments
- •In B2B, the strongest signal is revenue or explicit willingness-to-pay
- •Founding resembles sales: empathy, discovery, ROI framing, and stakeholder mapping
