What we know about AI is changing + me as an action figure

Dr. Ryan Ries here. This week, I want to discuss a few different articles that have come across my desk recently, and some observations about how people’s interactions with AI are evolving. 

If you're managing a team, making decisions about AI adoption, or just trying to stay ahead of where this technology is heading, these patterns are reshaping how work gets done, and you should keep up to date with what’s happening.

The Largest Study of ChatGPT Usage Ever

OpenAI just released an analysis of 1.5 million conversations from their 700 million weekly ChatGPT users. That's 10% of global adults using ChatGPT weekly. This is the most comprehensive look at actual AI usage we've ever had.

First, I found the demographics section super interesting. Early on, when genAI exploded, we saw a gender gap in usage. This has since nearly disappeared, with feminine names going from 37% of users in January 2024 to 52% by July 2025. ChatGPT has also seen explosive growth in low- and middle-income countries, with adoption rates over 4x higher than in wealthy nations. Access and democratization of information and technology have been huge benefits of this technology.

Now, taking a look at the insights on how we work:

Work usage is only 27% of total ChatGPT activity—and that percentage has actually dropped from 47% in June 2024. While work usage continues growing in absolute terms, consumer usage is exploding faster.

When people do use ChatGPT for work, writing dominates everything else. 40% of all work messages involve writing tasks: email drafting, document creation, and editing. Two-thirds of those writing requests are about editing existing content, not creating from scratch. 

The second biggest work use case is decision support, accounting for 58% of work activity. People are using AI for information gathering, problem-solving, and advice.

Here’s the link to the full study in case you’re interested in diving more into it.

Microsoft's Switches It Up

This next story did make me chuckle a little.

Microsoft, after investing billions in OpenAI, is now integrating Anthropic's Claude models into Office 365 Copilot.

Internal testing showed Claude outperformed OpenAI's GPT models in specific practical applications. Claude produced more accurate results for financial functions in Excel and created PowerPoint presentations that were cleaner and more visually appealing.

Microsoft isn't abandoning OpenAI. They remain the primary partner. But they're addressing customer frustrations with buggy Copilot features by bringing in models that perform better for specific use cases. They're paying Amazon Web Services (Anthropic's cloud host) for access while keeping Copilot at $30 per user per month.

This move also comes during tense restructuring negotiations with OpenAI, so Microsoft is reducing dependence on a single partner while strengthening its bargaining position.

The lesson here is that even massive AI partnerships can shift quickly based on performance data.

The Matrix Was Right About Power

Speaking of power dynamics, there's an uncomfortable parallel between The Matrix and today's AI race that I’ve been thinking about.

In the movie, the robots didn't take over because they were evil—they took over because they needed more power, and humans turned out to be an excellent energy source.

Today's AI development is creating the same insatiable hunger for energy. We're so focused on winning the AI arms race that we're literally choosing computational power over climate commitments. 

Just like in The Matrix, everyone’s pursuit of advanced AI is fully reshaping the world’s resources and energy. Anyone else seeing this as eerie? Just me?

Claude Users Are Going Full Autopilot

Anthropic's latest Economic Index report revealed something that should grab everyone's attention: the percentage of "directive" conversations—where users give Claude a task and let it run autonomously—jumped from 27% to 39% in just eight months.

People are shifting from "AI as assistant" to "AI as employee."

The breakdown is telling:

• Coding dominates at 36% of all tasks
• Educational tasks surged from 9.3% to 12.4%
• Business users are 44% more likely to use full automation compared to consumer users
• Geographic divide: wealthy countries use AI for collaboration, while developing nations lean into full automation

Many people using this technology have been striving for autonomous AI since day one. But as we know, quality control becomes much more challenging when 39% of interactions are fully autonomous. How do you ensure the AI is learning the right lessons from the right data? 

What This Means for Your Organization

Each of these stories points to the same reality: AI usage patterns are evolving faster than most organizations are adapting.

Questions you should be asking:

• How much autonomy are you comfortable giving your AI systems?
• Are you preparing for a workforce that expects AI to handle complete workflows?
• How diversified is your AI strategy if partnerships can shift this quickly?
• How are you balancing AI ambitions with resource constraints?

The organizations that thrive will be the ones that understand these usage patterns and adapt their strategies accordingly.

Let me know what you're seeing with AI usage in your organization. Are your teams pushing for more autonomy, and how are you managing that transition? Have new use cases cropped up in your organization with the latest shifts in this technology?

Until next time, 
Ryan

Now, time for our AI-generated image and the prompt I used to create it. I saw this prompt online and thought we should give it a whirl this week.