[1/2] McKinsey Reports on Gen AI (AI column)
I am continuing my meta-examination of various AI adoption reports, and I loved the McKinsey series of posts that systematically and rhythmically explore the issue. They started doing it in 2023 yearly 2024 and 2025This allows you to track trends and see how the situation develops in time. I will talk specifically about the following reports.
- 2023 year: The state of AI in 2023: Generative AI’s breakout year - survey 11 along 21 April 2023, 1684 Respondents, publication 1 August 2023 - 2024 year to start: The state of AI in early 2024: Gen AI adoption spikes and starts to generate value - survey 22 February 5 March 2024, 1363 Respondents, publication 30 May 2024 - 2024 mid-year: The state of AI: How organizations are rewiring to capture value - survey 16 along 31 July 2024, 1491 respondent, publication 12 March 2025 - 2025 year: The state of AI in 2025: Agents, innovation, and transformation - survey 25 June 29 July, 1993 respondent, publication 5 November 2025 year (I'm more specific. told results)
If you look at these reports, there are a number of metrics that can be perceived as time series data and evaluate trends.
First, let’s start with the metric of penetration and its breadth.
1) “AI use/adoption in at least one business function” - the main "through" KPI This is the most stable KPI in the entire series: the proportion of respondents who use AI in at least one function.
- 2023 - 55%
- 2024 (beginning) - 72%
- 2024 (middle) - 78%
- 2025 - 88% This metric can be called a penetration metric. (“penetration”)She answers the question “is there AI somewhere”, but does not distinguish between pilot vs. prod.
2) “Organizations regularly using gen AI in ≥1 business function” - second through KPI (but 2025- Issue Nov 2025 he's not accented)
- 2023 - 33%
- 2024 (beginning) - 65%
- 2024 (middle) - 71%
- 2025 NA - in the text of the report is not given a separate figure (Focus shifted to agents and AI overall)
3) “AI spreads across multiple functions” - width of application (partially comparable) With this KPI is more complicated: it is “the same in meaning”, but the wording in different reports is different. (≥2 functions>1 function ≥3 function).
- 2023 less 33Percentage of AI adopted in two or more functions
- 2024 (beginning) - 50% – AI adopted in two or more functions
- 2024 (middle) - NA
- 2025more 66% – AI is used in more than one function; and 50% in three or more functions. It seems that the authors of the report were less interested in the comparability of data from year to year, and more interested in how to measure the breadth of coverage, so the question of the number of functions with the adoption of AI gradually increased.
In continuation I’ll talk about evaluating the effects of AI/GenAI on companies and the concept of high performers that are better at implementing tools. I told you about it. hereBut in the continuation of the post, you will see how the criteria for high performance mutate over time.
#Engineering #AI #Metrics #Software #Productivity #Economics #Whitepaper