ARK Invest’s Joseph Soyegh argues that AI moved from chatbots to agents in 2025, with dramatically better task completion, stable subscription pricing, and sharply lower cost for fixed performance. He says this is driving rapid enterprise adoption, strong revenue growth at OpenAI and Anthropic, and a potential multitrillion-dollar shift in software and infrastructure spending.
Watch on YouTube ›Get the market thesis, key claims, assets, contradictions, and follow-up questions from any financial video — then unlock a version personalized to your portfolio, watchlist, and favorite speakers.
Joseph Soyegh, a research analyst on ARK Invest’s next generation internet team, presents the AI productivity section of Big Ideas 2026. His core message is that 2025 marked a transition from AI chatbots to AI agents, as reasoning models and supporting tools improved enough that agents moved from handling only a few-minute human tasks to tasks taking 30+ minutes by year-end. He emphasizes that the value per dollar of AI tools has risen sharply because prices for subscriptions and frontier-like access have remained relatively stable while capabilities improved. He cites an OpenAI study suggesting ChatGPT saves knowledge workers about 50 minutes per day after discounting for task success rates, implying meaningful daily value relative to a $20 monthly subscription. …
Near term, the actionable setup is continued upside in AI productivity beneficiaries if model capabilities keep improving and enterprise budgets keep shifting toward automation. The main tactical risk is that adoption remains narrow or reliability issues slow the conversion from demos to durable spend.
Over the next few months, the base case is broader enterprise deployment and faster AI software revenue growth, especially if token usage and ARR continue compounding. That view weakens if model progress plateaus or if buyers delay rollout after early pilots.
Structurally, the transcript argues that AI becomes a persistent productivity layer that redirects a growing share of knowledge-work spending into software and compute. The durable implication is a multi-year capex and platform cycle tied to model performance, pricing, and access to chips.
2025 was the year AI chatbots matured into AI agents, helped by more performant reasoning models and better tooling.
Directly stated as the framing for the presentation.
AI agents improved from reliably handling tasks of 5 to 6 minutes at the start of 2025 to more than 30 minutes by year-end.
Quantitative performance comparison given by the speaker.
ChatGPT saves an average knowledge worker about 50 minutes per day and effectively pays for itself after about half a day of use.
Speaker cites an OpenAI study and converts it into a dollar-value estimate.
Unlock the full claims, asset map, scores, related transcripts, follow-up questions, and AI chat — shaped around your portfolio, watchlist, favorite speakers, and risks.
