Skip to content
#AI

[2/2] McKinsey Reports on Gen AI (AI column)

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

Continue. story About McKinsey, let’s move on to risk and negative impact metrics. (There's data here, but again, it's hard to compare.)

4) “Experienced at least one negative consequence”

  • 2023 - NA
  • 2024 (beginning) - 44Percentage of organizations have experienced at least one negative impact from gen AI
  • 2024 (middle) - 47Percentage of organizations have experienced at least one negative impact from gen AI.
  • 2025 - 51Percentage of organizations using AI have experienced at least one negative impact (AI is general, not just gen AI.).

5) Inaccuracy as a key risk (partly numerically)

  • 2023 - NA
  • 2024 (beginning) a risk that is significantly more likely to be attempted than last year; and 25Percentage of respondents note negative consequences from inaccuracy of gen AI
  • 2024 (middle) - NA
  • 2025 - almost 33% of all respondents report consequences due to AI inaccuracy.

Value/EBIT, where the metric repeats, but thresholds and definitions float Here, McKinsey gives numbers, but in different editions it changes what we consider a success.

6) Share of companies with EBIT effect (thresholds)

  • 2023 - 23% of respondents say that ≥5% EBIT of their organizations attributable to AI (flat YoY at the time.)
  • 2024 (beginning) only 5.2% (46 and 876 respondents) "report that a meaningful share of their organizations’ EBIT can be attributed to their deployment of gen AI"
  • 2024 (middle) - 17% say that ≥5% EBIT attributable to gen AI>80% do not see a “tangible impact” on gen AI’s enterprise-level EBIT
  • 2025 - 39% report any EBIT impact on enterprise level from AI (And most of them do.5%)

7) Share of High performers in the sample (not 1b1but it seems to be on a scale.)

  • 2023 AI high performers are defined by the criterion> 20% EBIT attributable to gen AI.” There is no direct indication of their number, but the estimate can be taken from other questions, for example, about “reskill larger portions of the workforce”. 50 High performers and others 863 (estimation 5.4% high performers)
  • 2024 (beginning) - “gen AI high performers” = 46 from 876 (≈5.3%) respondents (criterion> 10% EBIT attributable to gen AI)
  • 2024 (middle) - NA (This report doesn't say anything about high performers.)
  • 2025 - “AI high performers” ≈ 6Percentage of respondents (Other criteria: EBIT impact ≥ 5% + “significant value”)

As a result, we see that the share of “tops” for self-reported value remains in order. 5–6%, but due to the change in criteria, this is not an accurate trend. If you look at 2023 year 23The percentage of respondents said 5% contribution of AI to EBITDA 2025 year-over-year 6The percentage of respondents who see this contribution. This means that the height of achievement of high performers becomes lower. (The impact on EBITDA is more realistic).

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