The state of AI in 2026: On the road to ROI

McKinsey & Company | Dan Tinkoff, Lieven Van der Veken, Michael Chui

A McKinsey Global Survey on the state of AI in 2026 reveals that forty percent of large organizations with annual revenues exceeding $1 billion are scaling AI agents, up from 27 percent last year. This rapid adoption of agentic coding tools is prompting 32 percent of surveyed companies to build software in-house rather than purchasing commercial products.

Despite this operational expansion, enterprise-level financial impact remains concentrated, with only 37 percent of respondents reporting earnings before interest and taxes (EBIT) gains. Individual productivity has surged. However, rising operating costs, including token expenses, now constrain usage for 20 percent of organizations. Looking ahead, 39 percent of respondents expect workforce declines in the coming year, although actual layoffs over the past year fell short of previous projections. To bridge this gap between individual efficiency and corporate returns, high-performing organizations are fundamentally redesigning workflows rather than merely inserting technology into existing processes.

Comment

The shift toward in-house software creation enabled by agentic coding tools directly mirrors the operational philosophy of the US Army Software Factory. By leveraging automated code generation, defence organisations can bypass the lengthy and rigid Federal Acquisition Regulation (FAR) procurement cycles typically required for commercial software. This capability allows military software units to rapidly prototype and deploy tactical applications directly to the field. Consequently, the traditional reliance on large defence contractors for minor software modifications is diminished.

This transition shifts the primary budgetary bottleneck from software licensing fees to cloud computing infrastructure and API token consumption. For programmes like the US Air Force's Advanced Battle Management System (ABMS), the marginal cost of scaling autonomous agents across distributed nodes will be determined by compute availability rather than vendor lock-in. Ultimately, the long-term sustainability of these custom-built systems will depend on whether the US Army Software Factory can maintain secure, sovereign code repositories without relying on traditional systems integrators like Lockheed Martin.

Strategic Question for Discussion

If the US Army Software Factory increasingly relies on agentic coding tools to bypass traditional procurement, how will the Department of Defense evaluate the software's cyber vulnerability and mission-assurance standards without the traditional verification pipelines provided by systems integrators like Lockheed Martin?

The trajectory indicates that the Pentagon will likely establish automated, continuous authority-to-operate (c-ATO) pipelines specifically tailored for AI-generated code. My assessment is that this shift will force a transition from static, document-heavy compliance reviews to real-time, algorithmic vulnerability scanning embedded directly within the US Army Software Factory's development environment. Ultimately, this evolution will redefine the role of traditional defence contractors from software builders to independent verification and validation authorities.

Share your assessment in the comments below.

To read in details click here