テンプラファイのレポートで、ナレッジワーカーの95%が、AIが生成した文書を、実務で使用できる状態にするために修正していることが判明

Templafy’s survey of 2,000 US and UK knowledge workers found that 90% use AI for business documents at least weekly, but 95% still need to revise AI-generated output and spend nearly four hours per week reviewing, correcting and validating it. Accuracy was the most cited issue (40%), while 48% said extensive review leaves little meaningful time savings. Templafy reported that agent-led document workflows cut median creation time from roughly two hours to eight minutes across more than 29,000 sessions, with usage among eligible users rising from 31% in December 2025 to 66% in August 2026.

Analysis

The investable implication is not incremental demand for foundation models; it is a shift in enterprise spend toward the control plane around them: permissions, retrieval, approved-content orchestration, auditability and workflow-specific agents. This favors platforms with embedded enterprise identity and productivity distribution—MSFT through M365/Copilot, GOOG through Workspace/Gemini, and CRM where customer-facing document generation can be tied to governed CRM data. The nearer-term revenue capture is likely to accrue to incumbent suites via premium-seat attach rates rather than to standalone document-AI vendors, whose feature sets remain vulnerable to bundling.

For MSFT and GOOG, the key risk is that low-quality outputs delay broad paid-seat expansion: CIOs can rationalize pilots but defer enterprise-wide deployments until accuracy and compliance controls are demonstrably better. Over the next 1-3 months, investor focus should be on management commentary around Copilot/Gemini retention, active usage after initial rollout, and whether customers move from general assistants to repeatable workflow deployments. A sustained gap between installed seats and active usage would pressure the assumption that AI monetization converts rapidly into high-margin recurring revenue.

CRM has a more targeted upside path because contextual generation is inherently more valuable when grounded in account, pipeline, service and contract data; however, that advantage only matters if customers permit agents to act across systems rather than merely draft content. The contrarian point is that better governance may reduce token consumption per task, limiting usage-based infrastructure upside even as software ROI improves. This is a low-impact vendor-sponsored data point rather than a standalone trading catalyst; the signal becomes actionable only if it coincides with disclosed acceleration in paid AI workflow adoption.

Key Decisions for Investors

• Maintain a 6-12 month preference for MSFT over GOOG within mega-cap AI: MSFT has the clearer enterprise monetization channel through existing M365 contracts and identity/workflow integration. Reassess if Copilot paid-seat growth or renewal commentary materially weakens for two reporting periods.
• Watch for a CRM long catalyst into the next earnings cycle only if management discloses rising Agentforce production deployments, data-cloud attach, or measurable service/sales workflow automation. Absent those KPIs, do not underwrite document-generation demand as incremental revenue.
• Use a relative-value expression—long MSFT / short a basket of smaller standalone enterprise-AI/document automation vendors—over 6-18 months where liquid names are available. Bundled productivity-suite functionality and procurement consolidation are the structural risk to point solutions; stop the trade if independent vendors show sustained net retention expansion above platform growth.
• Set an alert for enterprise AI disclosures that distinguish provisioned seats from weekly active users and completed workflows. A widening seat-to-usage gap would be a near-term negative for MSFT and GOOG AI revenue expectations and argues against chasing post-earnings AI multiple expansion.