What are the key takeaways from this a16z Podcast episode?

Key insights include: **Speed threshold:** Computer-using agents have surpassed average human speed and are approaching expert-level performance. This threshold matters because once agents outpace even power users, the default behavior shifts — handing any computer task to an agent becomes faster than doing it yourself, fundamentally changing how humans interact with software.; **Enterprise entry point — legacy systems:** The clearest near-term enterprise automation targets are ERPs, CRMs, and procurement workflows that lack APIs. One example: agents can replace a three-step human chain — team spreadsheet to ops software to vendor outreach — by handling the entire procurement cycle autonomously across software that hasn't changed in fifteen to twenty years.; **Guardian auto-approval framework:** OpenAI's Guardian system uses a secondary model to evaluate every action a computer-using agent takes, checking alignment with original intent before execution. Enterprises can configure which action types — downloads, new website visits, data transfers — require human approval, creating a tunable dial between full autonomy and step-by-step oversight.

What did Ari Weinstein and James Sun discuss on a16z Podcast?

OpenAI's Ari Weinstein and James Sun join a16z's Sima Amble and Fabrizio Serafini to examine how computer-using agents now operate faster than average humans, where enterprise and consumer automation is headed, and how companies can deploy agents across legacy software while maintaining audit controls and human oversight. Key topics include: **Speed threshold:** Computer-using agents have surpassed average human speed and are approaching expert-level performance. This threshold matters because once agents outpace even power users, the default behavior shifts — handing any computer task to an agent becomes faster than doing it yourself, fundamentally changing how humans interact with software.; **Enterprise entry point — legacy systems:** The clearest near-term enterprise automation targets are ERPs, CRMs, and procurement workflows that lack APIs. One example: agents can replace a three-step human chain — team spreadsheet to ops software to vendor outreach — by handling the entire procurement cycle autonomously across software that hasn't changed in fifteen to twenty years..

How long is this episode of a16z Podcast?

This episode is 45 minutes long. SignalCast provides an AI-generated summary so you can get the key insights in about 3 minutes.

a16z Podcast

OpenAI: When AI Agents Use Computers Faster Than You

October 26, 2026

45 min episode · 2 min read

Ari Weinstein,James Sun,Sima Amble

Episode

45 min

Read time

2 min

Topics

Startups, Fundraising & VC, Artificial Intelligence

AI-Generated Summary

Published Oct 26, 2026

Key Takeaways

✓Speed threshold: Computer-using agents have surpassed average human speed and are approaching expert-level performance. This threshold matters because once agents outpace even power users, the default behavior shifts — handing any computer task to an agent becomes faster than doing it yourself, fundamentally changing how humans interact with software.
✓Enterprise entry point — legacy systems: The clearest near-term enterprise automation targets are ERPs, CRMs, and procurement workflows that lack APIs. One example: agents can replace a three-step human chain — team spreadsheet to ops software to vendor outreach — by handling the entire procurement cycle autonomously across software that hasn't changed in fifteen to twenty years.
✓Guardian auto-approval framework: OpenAI's Guardian system uses a secondary model to evaluate every action a computer-using agent takes, checking alignment with original intent before execution. Enterprises can configure which action types — downloads, new website visits, data transfers — require human approval, creating a tunable dial between full autonomy and step-by-step oversight.
✓WebMCP as a collaboration standard: Rather than agents fully replacing human website interaction, WebMCP allows agents and humans to share control of the same interface. A practical workflow: the agent handles tedious steps like form-filling or data lookup, then hands control back to the human for decisions like payment confirmation, preserving intentional human touchpoints.
✓Startup opportunity in vertical diffusion gaps: AI adoption remains uneven across industries. Restaurant groups, for example, use computer-using agents to photograph back-of-house inventory, forecast next-day demand, auto-order via procurement software, and adjust staff schedules in an ERP — all in one workflow. Founders with deep vertical expertise can build end-to-end solutions where general-purpose tools haven't yet reached.

What It Covers

OpenAI's Ari Weinstein and James Sun join a16z's Sima Amble and Fabrizio Serafini to examine how computer-using agents now operate faster than average humans, where enterprise and consumer automation is headed, and how companies can deploy agents across legacy software while maintaining audit controls and human oversight.

Key Questions Answered

•Speed threshold: Computer-using agents have surpassed average human speed and are approaching expert-level performance. This threshold matters because once agents outpace even power users, the default behavior shifts — handing any computer task to an agent becomes faster than doing it yourself, fundamentally changing how humans interact with software.
•Enterprise entry point — legacy systems: The clearest near-term enterprise automation targets are ERPs, CRMs, and procurement workflows that lack APIs. One example: agents can replace a three-step human chain — team spreadsheet to ops software to vendor outreach — by handling the entire procurement cycle autonomously across software that hasn't changed in fifteen to twenty years.
•Guardian auto-approval framework: OpenAI's Guardian system uses a secondary model to evaluate every action a computer-using agent takes, checking alignment with original intent before execution. Enterprises can configure which action types — downloads, new website visits, data transfers — require human approval, creating a tunable dial between full autonomy and step-by-step oversight.
•WebMCP as a collaboration standard: Rather than agents fully replacing human website interaction, WebMCP allows agents and humans to share control of the same interface. A practical workflow: the agent handles tedious steps like form-filling or data lookup, then hands control back to the human for decisions like payment confirmation, preserving intentional human touchpoints.
•Startup opportunity in vertical diffusion gaps: AI adoption remains uneven across industries. Restaurant groups, for example, use computer-using agents to photograph back-of-house inventory, forecast next-day demand, auto-order via procurement software, and adjust staff schedules in an ERP — all in one workflow. Founders with deep vertical expertise can build end-to-end solutions where general-purpose tools haven't yet reached.

Notable Moment

Weinstein described a near-future shift where agents adapt to users rather than users adapting to agents — with proactive systems like Dots already catching missed calendar invites from external emails that never surface on internal calendars, flagging them without any user-initiated prompt.

Know someone who'd find this useful?

Episode Transcript

We've gotten computer use now to a place where it's faster than the average human, and I think we're getting really close where it is superhuman. Sorts of tasks do you think are ready to hand over to the agent versus the ones like, okay, it still requires some checking and correcting? Computer using agents now since Astra is now capable of doing anything that a person can do. We think a lot about, like, what new consumer experiences can we enable with a faster, better model. And then on the enterprise side, that's kind of naturally why the safety becomes more interesting. Do we give the ability to control the amount of queue in the loop there is and then the auditability? It's unlike any other thing that we've built where the developer has to do some work to adopt it. Someone has to build a connector, MCP server. You know, computers work with everything. Now we have, like, capable models. They're getting, like, cheap enough to run at scale. What should kind of companies think about automating? I think of it as in two big categories. The first category is, like AI agents can increasingly use the same software we do, clicking buttons, navigating websites, and completing work without dedicated APIs. What changes when they become faster at it than we are? A sixteen z's Sima Amble and Fabrizio Serafini joined OpenAI's Ari Weinstein and James Sun to explore the rapid progress in computer using agents and why computer use could become a universal interface for AI. They discussed how agents could automate tasks across legacy enterprise software, finance, procurement, and sales, alongside the technical advances driving progress, including why the harness surrounding a model can nearly double its speed. They also examine how enterprises can control and audit agents. And what comes next, AI that proactively works in the background, collaborates with us in real time, and eventually navigates software faster than expert humans. Welcome to the a 16 z podcast. I'm Sima Amble. I'm joined by my colleague, Fabrizio Serafini. In August, we wrote about computer using agents and how they're starting to do the real work inside companies. Today, I'm joined by Ari Weinstein and James Sun from OpenAI who are going to help unpack what is going on with computer using agents from their perspective and share some thoughts on the future. Ari, James, thanks for joining us. Of course. Happy to be here. Yeah. Thanks for joining us. Ari, kick it off and tell us a little bit about what in Astra has made computer using agents move forward and what's really unlocked. Yeah. So Astra's an amazing model. It's good at so many things. Our computer use models were already so good before Astra. You know, like, Soul itself is amazing at computer use. So Astra really helps with a few things around the edges in my sort of anecdotal experience. One is it is a little bit faster on average because …

Get the full transcript (9,412 words) + summary by email — free

One-time email with the complete transcript and AI summary of this episode. No account needed.

One email, no spam. We’ll also show you what SignalCast does.

Browse all a16z Podcast transcripts →