As of late 2025, there is clear movement toward AI-agent-based software, but not clear evidence that large, sales- and marketing-heavy enterprise software vendors have been materially cannibalized or that their traditional go-to-market motions have become unnecessary.

1. Large incumbent enterprise vendors remain dominant and sales-led. Salesforce is still the world’s largest enterprise applications vendor, generating tens of billions in annual revenue and maintaining a workforce of over 76,000 employees, with tens of thousands in sales and support roles, indicating that a classic sales-intensive GTM model remains central to its business. (en.wikipedia.org)

2. Incumbents are adding AI agents, not being displaced by agent-native competitors. Salesforce and other major enterprise software firms (e.g., Snowflake, ServiceNow) have launched their own AI agent platforms such as Agentforce and similar offerings, folding agents into their existing products and sales channels rather than being replaced by external AutoGPT-style competitors. (investors.com)

3. AI agent startups are growing but have not obviously cannibalized the large vendors. Startups like Sierra (customer-service agents), Decagon (AI support agents), Docket (AI sales engineer/seller), and Alta (AI SDR and GTM agents) have raised substantial capital and are scaling, with Sierra reportedly reaching a ~$10B valuation and >$100M ARR. But these numbers are still small relative to incumbents’ revenue bases, and coverage frames them as new competitors operating alongside — not yet hollowing out — giants like Salesforce. (sfchronicle.com)

4. Industry analyses describe agentic AI as emergent and largely ahead of its full impact. Gartner estimates that only about 5% of enterprise software includes AI agents today, with a forecast of 40% by 2026, explicitly portraying the agent transition as still in its early phases. Articles on “agentic AI” stress that many enterprises are not yet ready to rely heavily on autonomous agents and are focusing on narrow, supervised deployments, again implying that a broad structural upheaval has not yet played out. (blockchain.news)

5. AI is automating internal roles, but that’s not the same as external cannibalization of vendors or their GTM. Some companies, including Salesforce and Klarna, have reduced or reshaped support headcount by deploying AI agents, showing internal labor substitution. But this is different from AI-agent-based competitors making entire enterprise vendors or their sales organizations obsolete; the core vendors are, in fact, the ones deploying these agents. (businessinsider.com)

Because Chamath did not give a specific end date beyond a vague “matter of time” / “over the coming years,” and we are only about 2.5 years past the prediction, the medium-term horizon he seemed to imply has not clearly expired. At the same time, available evidence shows early competitive pressure and experimentation rather than the kind of large-scale cannibalization and obviation of traditional enterprise sales motions he envisioned.

Given the combination of (a) clear partial movement toward agents, but (b) ongoing dominance and sales-led GTM of incumbents, and (c) the absence of a firm deadline in the prediction, the fairest current verdict is that it is too early to definitively judge whether his forecast will ultimately prove right or wrong.