← @bgSponge

Notes on AI adoption in the enterprise (small sample size, very uninformed, but probably…

Summary

Slava Akhmechet argues enterprise AI adoption is currently limited to specialized vertical products like coding, while general tools like meeting summarization suffer from atrocious engagement metrics.

  1. General enterprise AI tools fail.

    Meeting transcription, email autocomplete, and document search show atrocious engagement metrics and provide little tangible value to the majority of employees.

    meeting transcription/summarization no one reads, email autocomplete no one uses

  2. Vertical products drive meaningful adoption.

    Real utility is concentrated in specialized categories like coding and first-line customer support, despite much experimentation resulting in abandoned slop.

    meaningful adoption (coding, first line customer support, etc.) is happening through specialized vertical products

  3. Shadow AI usage bypasses IT.

    A sizable group of young, forward-looking employees uses personal Claude or ChatGPT subscriptions to copy-paste work, a practice IT cannot easily block.

    IT departments can't shut it down because that would eliminate 98% of legitimate AI usage

  4. Compute demand will explode.

    Current inference demand comes from a few massive use cases, but Akhmechet predicts demand could rise 10^6x or 10^9x once models transform workflows.

    demand for compute will go up 10^6x or 10^9x

  5. Investor sentiment risks an AI winter.

    Frothy market sentiment creates a dangerous race against model capabilities; a stall in progress could trigger a massive correction and delay progress by a decade.

    we're in a race between investor sentiment and model capabilities

Read the original ↗Save to your Sponge

Summarized with Sponge. Quotes are the author’s own words.

More from @bg

All readings →