Prepare AI integration upstream
Why does AI produce spectacular results in some companies, and almost insignificant ones in others?

Antoine HeftlerCo-founder

Why does AI produce spectacular results in some companies, and almost insignificant ones in others? The latest report from OpenAI explains precisely what creates that gap. ⤵️
OpenAI's latest report explains precisely what creates that gap.
One pattern runs through the whole of the State of Enterprise AI 2025: AI creates value only when workflows, data, and responsibilities have been clarified upstream.
And the report goes further, showing how those gaps take shape. In the companies the report treats as "high-performing organizations": – teams use AI to standardize analysis, improve the quality of deliverables, and structure internal documentation; – repetitive tasks (summaries, first drafts, structuring files) are systematized, which explains the 40 minutes saved per day on average; – some teams (engineering, data, communications) gain 60 to 80 minutes through technical rewriting, log analysis, the creation of complex drafts, or the synthesis of data tables — uses that are impossible without stable workflows; – managers see better decisions, because the information is better prepared and more consistent.
By contrast, in companies that have not prepared, the report observes recurring patterns: – AI is used in isolation, often for simple text generation; – workflows stay unchanged, which mechanically limits the gains; – teams struggle to build on what they learn, because AI is not built into any repeatable process; – individual benefits never reach the scale of the organization.
The report does not directly quantify the impact gap between these two kinds of companies, but its macro analysis — international, cross-sector, and cross-function — yields a robust finding: "AI adoption alone is not enough." In other words, the organizations that prepare change the way they produce, document, and decide.
The question the report leaves between the lines — and that decision-makers now have to face — is therefore not "which tool should we choose?" but: "what have we prepared so that AI can actually work with us?"
💡 The report is available here: https://openai.com/index/the-state-of-enterprise-ai-2025-report/