AI summary by TwinMind
Pricing in the AI Era: From Inputs to Outcomes, with Paid CEO Manny Medina
Sequoia Capital
45m 30s
AI Agent Market Landscape - What's Working
- Hedgehog vs Fox analogy: Narrow specialization currently printing money
- Successful examples:
- Quandre: Policy renewals in insurance
- Expo: Penetration testing, replacing pentesting agencies
- Happy Robot: Calling truckers for freight brokers, spins up 2,000 agents to negotiate loads
- Owl: Insurance claims review
- Key pattern: Targeting narrow problems with large addressable populations, no clear software solutions, people-heavy processes
- Not replacing people directly: Replacing BPOs (Business Process Outsourcing companies)
- Broader scope challenges: AI SDRs too broad, creates market swirl
- Future potential: AI assistants/EAs not quite ready yet due to complexity (time zones, multiple business lines)
Market Transformation Thesis
- Disagreement with Kleiner hypothesis: AI won't start with highest-paying jobs (developers, lawyers, accountants, doctors)
- Manny's contrarian view: AI will succeed where jobs are hard to backfill
- Insurance actuaries, adjusters
- High-turnover BPO roles (6-month average tenure)
- Jobs people don't want to do long-term
- Market dynamics: Higher-paying jobs will attract too much competition (OpenAI, Anthropic going after lawyers/accountants)
- Dual approach theory:
- Co-pilot for creative/high-paying jobs (Harvey in legal, collaborative workflows)
- Full autopilot for lower-paying, less creative roles
Four Pricing Models That Work
- Activity-based pricing: Credit consumption model, easy to show work done
- Workflow-based pricing: String multiple activities together (e.g., document review), moves toward value-based pricing
- Outcome-based pricing: Charge for measurable results, includes outcome bonuses
- Pay-per-agent pricing:
- Replace human equivalent (SDR costs $70-90K/year fully loaded)
- Charge $20K per agent + performance bonuses
- Tap into headcount budget vs. tools budget (HR vs. RevOps)
Pricing Maturity Curve
- Bottom requirement: Must move beyond pure activity-based pricing to avoid commoditization
- Progression: Activity → Workflow → Value-based → Bespoke contracts
- Custom contracts enabled by AI: Can interpret complex deals, calculate annualized value, analyze contract bodies
- Customer default: Always choose easiest purchasing method first (fixed/consumption pricing)
- Vendor responsibility: AI agent builders must proactively move customers to value-aligned pricing
Cost Structure and Margins
- Token cost trajectory: Unlikely to decrease in reasoning era
- Inference-time compute more valuable than training-time compute
- Deeper thinking required, hallucination risks with workflow chains
- Better to have models do most work vs. small workflow boxes
- Multi-modal cost complexity:
- Beyond LLM costs: cloud, third-party APIs
- Avatar companies: phone time duration, dial volume
- Voice agents: call duration costs
- Margin visibility problem: Evil frameworks proxy all activity, can't track per-customer/per-agent profitability
- Value capture mismatch: Customers capturing all savings value, not sharing with AI agent providers
Paid Company Overview
- Mission: Business engine for AI companies
- Services: Billing, invoicing, monetization, pricing, collections, revenue recognition, vendor management, margin management
- Pricing guidance feature: Calculates human equivalent cost for same work by country/skill level
- Problem identification: Outreach experience showed existing software not built for evolving pricing models and outcome delivery
Vibe Coding and Development Speed
- Paid built in 1.5 months: Half using vibe coding
- Philosophy: Don't debug vibe code, throw away and restart
- Competitive advantage: Enables rapid iteration and customer conversations
- Market impact: Anyone can now "vibe code a ServiceNow"
Current Market Dynamics
- Vibe revenue phenomenon: Companies buying AI agents due to AI mandates, not proven ROI
- Renewal reckoning coming: Will separate successful companies from failures
- BPO counterattack expected: Large BPO companies will deploy their own agents using internal data
- Early stage energy: Founders enjoying discovery phase, innovation every 7 months with model improvements
Founder Advice and Philosophy
- Stay narrow: Focus on very specific customer segments
- Disregard TAM concerns: Small TAMs become big TAMs with superb execution
- Quality over breadth: Better to nail small audience than be mediocre for large audience
- Modern founder advantages: More sophisticated about company building, starting with better initial conditions
- Team culture: Fun-focused, serving an industry that can't believe they're getting paid to have fun
Lightning Round Insights
- Mount Rushmore founders: Jeff Bezos, Todd Olson (Pendo), Sam Altman, Collison brothers
- Essential reading: Rich Manning's statistical NLP book from Stanford
- Can't live without: Perplexity for everything, Claude as writing companion
- AGI timeline: Believes it's already here but underutilized
- Models commoditizing: No, not in reasoning era - input tokens 6-8x more expensive than predecessors
- AI optimism: Scaffold for human imagination, allows seeing farther and longer
