For all the hype surrounding artificial intelligence, the first large-scale analysis of how people currently use ChatGPT delivers a surprisingly pragmatic conclusion: AI creates the most value when it helps people think, decide, and communicate more clearly, not when it tries to replace them.
According to recent research based on OpenAI data and analyzed by the National Bureau of Economic Research, nearly 80 percent of ChatGPT usage falls into three categories:
- Practical guidance
- Writing
- Seeking information
Technical coding, experimentation, and self-expression together account for less than 20 percent.
For CIOs and business leaders, this data provides an initial snapshot and a potential direction. It explains why some organizations are extracting real value from AI, while others remain stuck in pilot mode. It offers a clear roadmap for how AI, Copilot, and ERP can work together to improve outcomes.
Currently, AI’s Value Lies in Decision Quality, Not Automation Volume
Much of the AI conversation in enterprise software centers on automation:
- Can AI replace users?
- Can it write code?
- Can it eliminate manual effort?
Those questions are valid and may be priorities, but there is a gap between the promised use of AI and its practical application.
The data shows that business users overwhelmingly turn to ChatGPT for cognitive support, not execution:
- Clarifying complex decisions
- Structuring ambiguous problems
- Improving Business Writing
- Synthesizing information across sources
In other words, AI is being used as a thinking amplifier.
Why This Matters for ERP Leaders
The thinking amplifier distinction matters deeply in ERP environments:
- During implementations where failures do not stem from missing functionality but from poor early decisions and inconsistent communication across the organization.
- During ongoing operations, the real value of AI is not being accessed.
ERP projects fail far more often due to:
- Unclear requirements
- Poor data readiness
- Misaligned expectations
- Weak executive project management
Software limitations are rarely the issue.
In fact, it has long been my view that if an organization can fix the issues above, it would be better off running a “classic” ERP environment, where “Excel Runs Production,” than implementing a modern cloud-based ERP platform with a 70% failure rate.
The primary deployments for AI in organizations appear to be:
- A chatbot for users
- A reporting assistant
- A helpdesk replacement
These capabilities are helpful, but they don’t address where AI can have a more significant impact in ERP.
The question for most business leaders is where to start applying AI in more powerful roles and in a logical manner.
Why Most Organizations Still Underperform with AI
If the value is so clear, why do so many AI initiatives stall? More importantly, why are the best AI use cases not even being contemplated?
The initial data suggests that most organizations use AI as an ad hoc tool.
More practical and productive AI use cases that are tried appear to launch:
- Without structure
- Without governance
- Without business-specific context
The result is inconsistent outputs, limited trust, and no repeatable value. AI becomes anecdotally impressive but not operational at scale.
This is especially problematic in ERP environments, where consistency and repeatability are prerequisites for scale.
What Should CIOs and Business Leaders Do Next
The question is: how do business leaders enhance the current predominant use of AI as a cognitive support tool to automate high-value workloads and improve decision quality?
The issue is that other research indicates most business leaders don’t know where to start a high-value AI project.
One-off consulting engagements that rely on the expertise of a specific consultant or consulting organization will be a hit-or-miss proposition, depending on the experience and bias of that organization or individual.
This is where AI and a large dataset come to the rescue, introducing:
The HandsFree Roadmap to AI Agent Projects with AI Agent Assessments
