Conviction founder Sarah Guo posted on Oct 27 that enterprise workflow automation is "in the autocomplete / early claude code era", "but not for long", drawing about 80K views and nearly 500 likes. Box CEO Aaron Levie quoted it the next day and argued that most enterprise workflows have not had agents applied in any meaningful way, putting deployment at probably under 1%, mainly because of AI capability itself; he says computer use, long-running agents and agent swarms are changing that. His post has about 635 likes, 486 bookmarks, 63K views and 150+ replies. These are opinions and rough estimates, not measured statistics.
Key Takeaways
- ✓Sarah Guo (Conviction founder): enterprise workflow automation is still in the "autocomplete / early claude code era", but not for long
- ✓Aaron Levie (Box CEO): most enterprise workflows have not had agents applied meaningfully, and deployment across the economy is "probably < 1%", which is his estimate
- ✓Levie says the main bottleneck is AI capability itself, with enterprise adoption speed a smaller factor
- ✓He points to computer use, long-running agents, harness improvements and agent swarms as what now makes more complex workflows workable for agents
- ✓His view: the shift is less about users prompting agents and more about organizations re-engineering workflows, with the biggest impact in finance, life sciences, healthcare and other information-heavy fields
Key Decision Metrics at a Glance
Heavy Claude Code use: compare subscription limits and API bills
Compare 40 dev plans & simulate token costs vs $20/mo subscriptions
In-Depth Technical Analysis
What happened
At 21:27 UTC on Oct 27, Sarah Guo, founder of Conviction and host of the No Priors podcast, posted on X: "we remain in the autocomplete / early claude code era of enterprise workflow automation, but not for long." At the time of writing the post had about 80K views, 486 likes and 100 replies.
At 04:51 UTC on Oct 28, Box CEO Aaron Levie quoted it with a longer post agreeing that the vast majority of enterprise workflows have not had agents applied to them in any meaningful way, and that "the economy probably has this deployed < 1% right now." His post had about 635 likes, 50 reposts, 486 bookmarks, 152 replies and roughly 63K views; the high bookmark count suggests many readers saved it as a reference.
To be clear, the under-1% figure is Levie's personal estimate; his post gives no data source for it.
Levie's argument
- Today: at best a company has a few systems that make a single LLM pass to make a decision or give feedback, which is not real automation.
- Why: partly the usual slow pace of enterprise technology adoption, but mostly, he argues, the capability of AI itself.
- What changed: computer use, long-running agents, harness improvements and surrounding infrastructure now let agents take on more complex workflows.
- Next: self-improving systems that take a workflow goal and raise execution success rates over many instances, plus agent swarms that let a company point any amount of compute at a problem.
- Where it lands: less about end users prompting agents for one task, more about organizations re-engineering the underlying workflows around agents, with the biggest impact in finance, life sciences, healthcare and other information-heavy fields.
What people are debating
Together the two posts have about 140K views, over 1,100 likes, 600+ bookmarks and 250+ replies. The open questions are whether the under-1% estimate holds up, and whether the bottleneck is model capability or enterprises' own adoption speed and process redesign (Levie himself puts it mostly on the former). For developers, the practical takeaway is that the enterprise opportunity is more about handing a complete workflow to agents and measuring its success rate over time than about building another chat interface.
Benchmark side-by-side against alternatives, or calculate monthly token cost vs subscription break-even.
Discussion & Comments
0Sign in to join the discussion
Connect with AI developers to exchange benchmark insights.