AI in 2026: Scaling Tools, Seeking ROI

Organizations are rapidly scaling AI and agentic tools, yet enterprise-level financial impact remains flat for the majority of businesses. While individual productivity has soared and companies are increasingly building their own software, operating costs are starting to limit usage. The path to significant ROI requires a fundamental redesign of workflows rather than simply inserting AI into existing models.

Key Points

• AI adoption is shifting toward enterprise-scale scaling, with large organizations leading in the use of autonomous agents.
• Agentic coding tools are enabling companies to build custom software in-house, disrupting traditional software procurement models.
• There is a disconnect between high individual productivity gains and stagnant enterprise-wide financial ROI.
• AI operating costs, including token expenses, are emerging as a constraint for one-fifth of organizations.
• High-performing companies achieve superior results by fundamentally redesigning workflows and prioritizing innovation over simple efficiency.

Sentiment

The community leans skeptical toward the article's optimistic enterprise-AI framing, especially around ROI and consultant credibility. It does not reject AI usefulness outright: many commenters accept that coding tools can help individuals, support small internal tools, and improve software work. The dominant mood is cautious and critical, with agreement on adoption but disagreement over whether adoption currently adds durable enterprise value.

In Agreement

• Several commenters agree with the article that organizations are increasingly using AI coding tools to build internal software rather than buying every SaaS product or feature.
• Some argue that small, department-level custom tools can be more effective than broad general-purpose software and that this pattern echoes earlier successful uses of spreadsheets, Access, SharePoint, Notes, and low-code systems.
• Supporters of the productivity thesis say AI is already obviously useful for individual work, code understanding, test generation, and replacing or rewriting legacy implementations.
• More nuanced commenters agree that weak enterprise financial metrics do not prove AI has no ROI, because benefits may diffuse unevenly, appear outside direct profit measures, or lag behind informal employee adoption.
• Some participants agree with the article's implication that companies that redesign workflows and measure impact well may gain an advantage over firms that merely sprinkle AI over existing processes.

Opposed

• Many commenters doubt that the article shows meaningful enterprise ROI, treating the report as consultant optimism around benefits that remain hard to prove.
• A prominent concern is that AI-generated internal software may repeat the problems of past end-user computing waves, producing fragile systems that later require professional cleanup.
• Skeptics argue that LLM-written code can be bloated, overengineered, underspecified, and hard to maintain, especially when business logic is undocumented or hidden in scattered scripts.
• Several commenters distrust McKinsey as a source and see the report as part of a broader pattern of management-consulting narratives that encourage questionable executive decisions.
• Some distinguish individual task acceleration from enterprise impact, arguing that faster coding does not necessarily shorten delivery timelines or improve profit when extra time is absorbed by verification, quality work, or organizational bottlenecks.