Delivering a Copilot license to each employee, without considering what tasks they perform or what specific needs they face, does not guarantee any type of transformation. A task-based adoption model starts with understanding the actual work people do, identifying where AI can have an impact, and building specific solutions. At the same time, employees who receive these tools without focused training tend to revert to their traditional methods, practically ignoring the AI that is provided to them.
If companies want artificial intelligence to truly transform work, they need to start with the basics: understand the tasks, identify critical processes, design integrations that respond to real needs, and accompany the technology with practical, not generic, training.
Today we are seeing a troubling trend: companies around the world are implementing artificial intelligence en masse, purchasing thousands of licenses for tools like Microsoft Copilot, without a true understanding of how this technology should be integrated into daily work. The approach, in many cases, has simply been to "put AI everywhere" and hope that people figure out how to be more productive.
To know how an assistant like Copilot can make a worker more efficient, one must first precisely understand what tasks they perform day-to-day. Otherwise, it may end up generating frustration and wasting resources. The problem is that not all workers use the same tools or face the same challenges. It is a strategy that confuses movement with progress.
Even among those who do use these tools intensively, the differences are enormous: the way a financial analyst uses Excel is very different from how someone in marketing or logistics would use it. Meanwhile, companies like Morgan Stanley or Walmart have demonstrated that the real opportunity of AI lies not in mass adoption without purpose, but in tackling specific and measurable problems.
Otherwise, Copilot and its equivalents risk becoming the corporate gym membership of 2025: everyone will have it, but very few will use it to change anything. Artificial intelligence should not be everywhere. And the truth is that very few organizations have that visibility. The numbers begin to reflect this disconnection.
To assume that a data engineer, a human resources analyst, and an operations manager will use AI in the same way is to completely misunderstand the reality of modern work. The "buy licenses for everyone" model stems from a blind faith that technology, by itself, will solve problems that haven't even been clearly defined.
This confusion is compounded by another challenge: many of the companies promoting these solutions, including Microsoft, also do not provide clear guidance on how to capture real value. According to Gartner, less than 6% of companies report having obtained significant return on investment from their Copilot deployments, and a growing proportion is delaying their implementations due to security concerns and low adoption levels.
Someone who rarely opens Word or Excel will hardly see relevant benefits from having a writing or analysis assistant. Morgan Stanley, for example, implemented a tool that automatically generates meeting notes for clients, integrating them into Salesforce and saving their financial advisors up to an hour of manual work daily. It should be where it truly solves something.
In both cases, the AI was designed to address a clear bottleneck, blending almost invisibly into the natural workflow. The difference is profound. Walmart used AI so that its employees could check inventory status in seconds and avoid stockouts.
