Skip to content
back to the archive page
#AI

[2/2] McKinsey’s Reports on Generative AI (Category AI)

#AI #Engineering #Metrics #Software #Productivity #Economics #Whitepaper

Continuing the discussion of McKinsey’s research, let us turn to risk and negative-consequence metrics. There are figures here too, but they remain difficult to compare.

4) “Experienced at least one negative consequence”

  • 2023: N/A.
  • Early 2024: 44% of organizations experienced at least one negative consequence from generative AI.
  • Mid-2024: 47% experienced at least one negative consequence from generative AI.
  • 2025: 51% of organizations using AI experienced at least one negative consequence. This now covers AI overall, not just generative AI.

5) Inaccuracy as a key risk: some numerical evidence

  • 2023: N/A.
  • Early 2024: significantly more organizations were trying to mitigate this risk than in the previous year, and almost 25% of respondents reported negative consequences specifically from generative-AI inaccuracy.
  • Mid-2024: N/A.
  • 2025: almost 33% of all respondents reported consequences from AI inaccuracy.

🤑 Value and EBIT: recurring metrics, changing thresholds McKinsey supplies numbers, but what counts as success changes between editions.

6) Companies reporting an EBIT effect: different thresholds

  • 2023: 23% said at least 5% of their organizations’ EBIT was attributable to AI, unchanged year over year at the time.
  • Early 2024: only 5.2%, or 46 out of 876 respondents, reported that a meaningful share of their organizations’ EBIT could be attributed to generative-AI adoption.
  • Mid-2024: 17% said at least 5% of EBIT was attributable to generative AI, while more than 80% saw no tangible effect on enterprise-level EBIT from it.
  • 2025: 39% reported some enterprise-level EBIT impact from AI; for most, it was less than 5%.

7) High performers’ share of the sample: similar scale, different definitions

  • 2023: the criterion was more than 20% of EBIT attributable to generative AI. Their number was not stated directly, but other questions offer an estimate. For example, the workforce-reskilling question had 50 high performers and 863 other respondents, giving an estimated high-performer share of 5.4%.
  • Early 2024: 46 of 876 respondents, about 5.3%, were “gen AI high performers,” defined as more than 10% of EBIT attributable to generative AI.
  • Mid-2024: N/A. That report did not discuss high performers.
  • 2025: “AI high performers” made up about 6%, with a different criterion: an EBIT impact of at least 5% plus “significant value.”

The top group by self-reported value therefore remains around 5–6%, but changing criteria mean this is not a precise trend. Looking back at 2023, when 23% reported a 5% AI contribution to EBITDA, the 2025 figure falls to 6% reporting such a contribution. My reading is that the bar for high-performer achievements has become lower, or estimates of AI’s EBITDA contribution have become more realistic.

#Engineering #AI #Metrics #Software #Productivity #Economics #Whitepaper