Friedberg’s claim is that over the coming years “essentially all” software-dependent business models rooted in traditional information retrieval would have to be rewritten to incorporate AI-driven synthesis.

Evidence through late 2025 shows:

• Widespread but not universal adoption: Surveys find ~72–78% of organizations use AI in at least one function and ~65–71% use generative AI, indicating rapid uptake but not near‑total penetration, and usually in specific functions rather than full business‑model rewrites. (mckinsey.com)
• Enterprise software not yet structurally transformed: Gartner estimates less than 10% of enterprise software was multimodal/GenAI‑infused in 2024, projecting 80% by 2030, which explicitly frames deep AI integration as a transformation still mostly in the future. (gartner.com)
• Core information‑retrieval businesses remain largely intact: In web search, Google still holds ~89–91% share in 2025; AI‑native or AI‑centric search tools (ChatGPT, Perplexity, etc.) represent only about 5–10% of queries, and Google’s business model is still primarily ad‑driven search, albeit with AI overlays. (infront.com) This suggests IR‑based models have been augmented, not yet fundamentally replaced en masse.
• Most companies are still experimenting, not fully restructured: Gartner, BCG, MIT and others report that although many firms are piloting or deploying GenAI, only a small minority achieve strong, measurable business value; many AI projects are siloed, underperforming, or stalled due to integration challenges and legacy systems. (businessinsider.com)

These data show a clear direction of travel toward AI-driven synthesis inside software and workflows, consistent with the spirit of the prediction, but they fall far short (so far) of “essentially all” IR‑based software business models being restructured or replaced. Because Friedberg’s horizon was “over the coming years” and major analysts themselves place the bulk of this structural shift closer to 2030, it is too early to decisively judge the prediction as either right or wrong.

Hence, the status as of November 2025 is inconclusive (too early).