238社が選ぶ「現場が支持する」AI 、Claude・NotebookLM が上位

In June 2026, ITmedia surveyed 238 companies on their priorities when selecting IT services and AI tools. The findings show a nuanced picture of how organizations now evaluate AI in practice, with priorities shifting based on company size and business context.

Across all respondents, the top three selection criteria when choosing IT services—including AI—were cost-effectiveness (53.4%), ease of use and operational simplicity (50.8%), and compatibility with existing tools and workflows (41.2%). However, when broken down by company size, the order of priorities shifts significantly. Large enterprises, constrained by compliance and governance requirements, weight security and vendor reliability (50.8%) and the ability to operate tools safely within existing systems (41.2%) as critical factors, whereas small and mid-size firms remain more focused on cost and implementation simplicity.

When asked which specific AI services they are using or considering, respondents cited Claude (Anthropic's LLM), NotebookLM (Google's tool for generating Q&A documents and wiki-style summaries), Microsoft Copilot, Gemini (Google Workspace integration), and Genspark. Claude received particular mention for the code-writing capabilities of Claude Code and its compatibility with development tools like Cursor. Respondents emphasized that they value Claude's ability to understand complex code problems and support multiple types of code-related tasks. NotebookLM was cited for its practical use in creating reference documents and learning materials. One respondent noted that when choosing between solutions like NotebookLM and Claude, or between Copilot and Gemini, the decision hinged on specific use cases rather than general technical superiority. Others described adopting AI tools for document Q&A, content creation, and process automation—tasks where AI delivers immediate, measurable productivity gains.

The survey also probed barriers to adoption of new AI services. Across all respondents, 37.0% cited "unclear implementation or operational hurdles" as a significant concern; 34.0% named "unclear fit with existing systems"; and 33.6% flagged security concerns. The barriers vary by company size and decision-making culture. Large enterprises focus heavily on security and change-management risk, asking whether tools can be operated safely in their production environments and whether vendors meet compliance standards. Small and mid-size firms, by contrast, worry more about whether they have the personnel and budget to integrate a new tool smoothly. These differences suggest that a one-size-fits-all AI adoption strategy does not work; organizations must tailor evaluation and rollout to their own constraints, governance maturity, and risk tolerance.

The survey reveals a critical shift in how organizations evaluate AI tools. Early adoption of AI services was driven by technical capability and novelty; the 238 respondents' feedback shows that decision-making has matured into a pragmatic assessment of alignment with specific business workflows. The top selection criteria—cost, ease of use, integration, security, and vendor trust—reflect real operational constraints rather than aspirational technology metrics.

Company size emerges as a determining factor in priority weighting. Small and mid-size firms, constrained by budget and personnel, favor lower total cost of ownership and tools that integrate cleanly with existing systems. Large enterprises, by contrast, face governance and compliance demands that make security certification, vendor reliability, and operational transparency non-negotiable. This split suggests that a single "best AI tool" does not exist; rather, organizations are matching tools to their own maturity and risk appetite.

The prominence of Claude, NotebookLM, and Microsoft Copilot in responses underscores that real adoption follows demonstrated value in narrow, repeatable tasks—code generation, document Q&A, content drafting—rather than aspirational general-purpose AI. Respondents frequently compared tools (Claude versus NotebookLM, Copilot versus Gemini) within specific use cases, signaling that the market is segmenting not by vendor but by application. The largest cited barriers—implementation ambiguity (37%), system fit uncertainty (34%), and security concerns (33.6%)—point to a gap between tool capability and organizational readiness, suggesting that future differentiation will belong to vendors who simplify onboarding and governance, not merely improve model performance.

No comments yet. Be the first to share your thoughts!