Available evidence by November 30, 2025 shows strong movement in the direction of Sacks’s prediction, but it does not clearly confirm that “a large share” of customer-support inquiries overall have already “effectively disappeared” due to in‑product AI assistants.

What has clearly happened (supports the direction of the prediction):

• Major SaaS and productivity vendors (e.g., Microsoft 365 Copilot, Google Workspace Duet/Gemini, Salesforce Einstein Copilot, HubSpot AI, Intercom Fin, Zendesk AI agents) have embedded AI assistants directly into their products to answer help questions, draft responses, and resolve common issues in‑product rather than routing to humans or separate ticket portals. These deployments are specifically marketed as reducing escalations to human agents and deflecting tickets.
• Case studies and vendor reports (from companies like Intercom, Zendesk, Ada, Forethought, and others) describe substantial ticket deflection (often 20–40%+ of incoming requests for certain workflows or problem types) when AI chatbots/assistants are deployed inside the product experience, reducing contact volume that reaches human support.
• Many startups now build products with AI help directly in the UI (tooltips, chat sidebars, command palettes) so users get real‑time, contextual assistance rather than filing tickets at all. This matches the mechanism Sacks described: “assistant gets built into the tool directly, so you never get to the point” of contacting support.

Why the overall truth value is ambiguous rather than clearly right or wrong:

• There is no broad, independent, industry‑wide data yet showing that across software products as a whole a “large share” of support inquiries have already disappeared. The best evidence is fragmented vendor case studies and marketing claims, which are not a reliable aggregate measure.
• Impact is highly uneven by sector and company size. Large tech/SaaS firms and AI‑forward startups have implemented substantial in‑product AI support; many traditional enterprises, regulated industries, and smaller vendors still rely heavily on human agents and conventional ticket systems.
• Overall customer support employment and spending have not visibly collapsed. While some organizations report efficiency gains and slower headcount growth in support, there isn’t strong macro‑level evidence that human‑handled support volume has already dropped by the kind of “large share” implied by the prediction.
• The prediction is somewhat qualitative (“a lot… just go away,” “large share”), so judging whether the threshold has been met requires a quantitative benchmark that the public evidence does not supply.

Conclusion

By late 2025, product‑embedded AI assistants are clearly handling many more customer questions in‑product and reducing the need for traditional support interactions in a growing portion of the software industry. However, because the scale of this effect across the whole market cannot be reliably measured and varies widely by company and sector, we cannot definitively say that a large share of all such inquiries has already ‘effectively disappeared.’ The claim is therefore best scored as ambiguous rather than clearly right or wrong based on currently available evidence.