23 Daily — On November 8, 2026, venture expert Jason Lemkin shared key decisions and trends in the AI and B2B fields during his 20VC x SaaStr session, including a 72% cost reduction in owned AI applications, approaches to evaluating total addressable market (TAM) versus market share growth, and recent valuation and acquisition activities in leading AI companies. This analysis is based on reporting from saastr.com.

Significant Model Cost Reduction Drives Pricing and Margin Strategies

Jason Lemkin revealed that by migrating their SaaStr Connect AI matching and recruiting application to Haiku version 5.5 paired with a cheaper model mix, their workflow costs fell by 72% while improving speed and output quality. This marked the first time since early 2025 that they observed such a sharp drop in inference costs in their actual billing. Lemkin emphasized that founders who do not routinely re-benchmark their model mix may be overpaying, while competitors who adjust pricing can maintain healthier margins. He advises teams to reassess production workflows with real use cases whenever new models launch to capitalize on cost improvements.

Growth Metrics Require Multi-Dimensional Interpretation

Taking Anthropic as an example, its GAAP revenue reportedly grew from $5 billion to

7 billion over the first three quarters of 2026. Although quarter-over-quarter growth rates reduced from 100% to 50%, the actual dollar amount of new revenue remained roughly constant, supporting a fivefold year-over-year increase. Lemkin notes that declining growth percentages are a natural effect of scaling. He recommends that startups present both net new annual recurring revenue (ARR) and growth percentages together to better communicate true expansion to boards and investors.

Pricing Pressure Forces Shift Toward Outcome-Based Models

Lemkin attributes Anthropic's growth deceleration to price competition, particularly as OpenAI sells similar products at about 30% cheaper and open-weight models lower market prices further. He advises AI companies to anticipate a 30-40% price drop within a year if pricing is mainly based on model capability. Pricing should shift toward the value of customer outcomes and workflows, as these factors retain value better than token-based cost reductions.

Agent Technologies Lower Use Barriers, Challenge Middlemen

Jason mentions operating 21 agents that mostly use programmatic tools like Higgsfield and Opus. DoorDash has incorporated a multichannel platform (MCP) integration to accommodate agent interactions. Founders should examine their product's top five common tasks to determine if agents can handle them via APIs or MCP. Any tasks still requiring human UI interaction risk being routed by agents to alternative services, potentially undermining marketplace fees.

Editor's take: Validating Market Expansion and Cost-Driven Pricing Strategies

Jason Lemkin highlights the importance for founders in AI and B2B sectors to continuously test whether their efforts are expanding TAM or just capturing existing market share, which shapes long-term growth strategies. Also key is experimenting with pricing approaches as AI cost structures evolve, moving from model capability markups to outcome-based pricing. Concrete verifications include tracking absolute new ARR growth, routinely benchmarking model costs against real workloads, and studying agent impact on user interaction. These insights provide valuable guidance for startups and investors in the dynamic AI market.