Key Takeaways
- Nearly a Third Have Walked Away from Software Purchases
- Productivity Rose, But Costs Stayed Flat
- Operating Costs as a Design Constraint
- One Thing to Check This Week
A survey released this week by McKinsey found that large enterprises are scaling agentic AI faster than smaller companies, and usage is growing quickly across the board. The survey covered more than 1,700 employees across multiple industries and company sizes. Perhaps the most notable finding is a shift in purchasing decisions. The report noted that some companies scaling agentic AI are replacing portions of their software purchases with coding tools. Nearly a third of respondents said they had decided not to buy at least one software capability because they could now build it in-house.
Productivity Rose, But Cost-Savings Responses Stayed the Same
Four out of five respondents in the report said AI had improved their personal productivity. However, financial results at the company level have yet to follow suit, and the share of employees reporting cost savings from AI was unchanged from last year. Operating cost pressures also showed up clearly in the survey. One in five respondents said operating costs were limiting how much their organization could use AI. At the same time, 28% said they were spending more than 10% of their overall IT budget on AI and expected that share to grow further next year. On the workforce side, while there have been layoffs in the tech industry this year, overall job numbers in the sector have held steady, and the report noted that the labor market is increasingly favoring workers who know how to work with AI. Two-thirds of companies, meanwhile, reported little to no change related to AI.
Advice to Re-Draw the Line Between Buying and Building
Michael Chui, a senior fellow at QuantumBlack, AI by McKinsey, and co-author of the report, said agentic tools consume more computing resources than earlier tools and therefore cost more, but that technology leaders are finding value in these use cases. He said organizations are developing new discipline to optimize the ROI of AI spending while still planning to increase investment. He added that spending on AI tools used broadly across an organization, such as chatbots, is now being managed as a standard cost of doing business, similar to office productivity and communication tools, and that agentic coding tools are creating a more viable option for bringing parts of software development in-house. Chui explained that this shift is part of a broader change in which developers' roles are moving from writing code directly to managing the output produced by AI.
Lieven Van der Veken, a senior partner at McKinsey, said that managing IT spending optimally requires deliberately deciding when to build versus when to buy, and taking greater ownership of a company's own technology and transformation agenda. He added that operating costs should be treated as a design constraint rather than an afterthought, and that this goes hand in hand with reshaping workflows as deeply as needed to realize the intended value.
Implications for AX at Korean Companies
While this survey does not break out responses specifically from Korea, there are three takeaways Korean companies can apply right away. First, reconsider the baseline for purchasing decisions. When reviewing a license renewal or a new software adoption, it's worth checking whether there is a process in place, before signing any contract, to determine whether the capability could be built in-house and who would be responsible for maintaining it afterward. Second, the advice to treat operating costs as a design constraint can be translated into practice by setting a cap on computing costs at the design stage, rather than after the adoption decision has already been made. Third, productivity and financial performance are not the same metric. In this survey as well, four out of five respondents reported improved personal productivity, while the share reporting cost savings remained unchanged from the previous year. If there is one thing worth checking in your own organization this week, it's whether internal AI performance reports list productivity metrics and cost metrics separately, and whether cost metrics are actually being measured at all.
Source: Enterprises bet on agents to build in-house software, boost productivity