We're thrilled to share this guest post from Andrew Jacobs, a leading L&D strategist and researcher who's spent the last decade helping enterprise organisations rethink learning in the age of AI. You may have already encountered Andrew's thinking if you've attended one of our Thinqi Roundtable events—he's a regular contributor and brings exactly the kind of grounded, research-backed perspective we know resonates with our community.
In this post, Andrew unpacks new research on how millions of people are actually using ChatGPT, and what it means if you're responsible for learning strategy in your organisation. Spoiler: the implications go way deeper than "we need a policy."
If you want to explore these ideas further with Andrew in conversation, watch our Thinqi Roundtable on demand where we dig into AI governance, the future of the L&D function, and what this all means for your strategy. We've also got upcoming podcast episodes with Andrew coming soon—where we'll move from insight to impact.
For now, read on—and if this resonates, you'll want to hear more.
How people are using ChatGPT and what it means for L&D
The new National Bureau of Economic Research (NBER) working paper How People Use ChatGPT (Chatterji et al, 2025) provides the first large-scale evidence of how millions of people actually use generative AI. It’s the clearest signal yet that learning at work has changed.
As L&D leaders, we can’t afford to treat this as an edge case. The data shows how ChatGPT has already become a global tutor, editor, and advisor. The implications are immediate for how we design, measure, and lead learning strategy.
The headline findings
- By July 2025, ChatGPT had 700 million weekly active users, sending 2.5 billion messages every day. Around 10% of the global adult population is using it regularly.
- Non-work dominates: Over 70% of use is personal, not professional.
- Writing is king at work: Emails, reports, summaries, translations. Two-thirds of this writing work is editing or improving human text, not creating from scratch.
- Decision support is the core value: Almost half of all messages are people Asking for advice or clarification. At work, Asking is growing faster than Doing, and is rated more useful.
- Tutoring is real: 10% of all use is tutoring or teaching. Education and coaching are already being outsourced to AI.
- Work use converges on two things: 81% of work-related interactions are about either (1) obtaining, documenting, and interpreting information, or (2) making decisions, solving problems, and thinking creatively.
- Demographics are shifting: Early gender and income gaps have closed. Younger people dominate overall usage but older, more educated professionals are more likely to use it for work.
- Global adoption is fastest in lower-income countries.
Why this matters for L&D
From content delivery to decision support
Most L&D functions still push content. The paper shows workers are turning to AI for guidance and judgment, not just outputs. Our role must shift to building capability to question, interpret, and apply AI-driven insight. This is where performance happens.
Writing skills are being redefined
If writing is the most common workplace use, L&D must rethink communication skills. It’s no longer about grammar and templates. It’s about prompting, editing, and adapting AI drafts into credible business outputs. That is a new skill set.
Self-determination has replaced self-direction
This isn’t just self-directed learning within organisational pathways. It’s self-determined. People are choosing what, how, and why they learn with AI on their own terms. Workout plans, coding tutorials, writing feedback, language practice. L&D no longer sets the menu — people do. Our role is to enable them to learn well, not to funnel them into our pathways.
Tutoring is already outsourced
With 10% of usage being tutoring, the “teacher” function of L&D is being bypassed. AI explains, drills, and coaches on demand. That doesn’t mean L&D disappears. It means our value is in integrating, validating, and connecting AI-driven learning to workplace performance.
Global equity is both opportunity and risk
AI is spreading faster in lower and middle-income countries. That can reduce barriers to learning but also risks embedding bias, poor quality, and unsafe practice. L&D has to take responsibility for governance, equity, and standards in how AI is used for learning.
Measurement has to change
Traditional metrics (completions, logins, satisfaction) mean little here. The gains are in speed, decision quality, and creative problem solving. L&D evaluation must adapt to track where AI improves performance and outcomes, not just where people consume content.
What to do next
- Focus on capability, not courses: Build AI literacy, prompting skills, and critical sense-making.
- Redesign communication training: Emphasise editing, adapting, and quality control of AI outputs.
- Support self-determined learning: Help people learn with AI responsibly, rather than trying to control the process.
- Integrate AI into performance support: Accept that AI is already the first-choice tutor and advisor.
- Lead on governance: Set the guardrails for safe and equitable AI use at work.
- Update evaluation: Prove impact in decision quality, speed, and outcomes, not in seat time.
The bottom line
The study shows ChatGPT is less about replacing jobs and more about reshaping knowledge work. It is already the co-pilot for writing, decision-making, and personal learning.
For L&D leaders, the task is clear: stop defending legacy pathways and start enabling AI-augmented performance, judgment, and self-determined learning.
That is the future of learning strategy.
About Andrew Jacobs
Andrew is an independent L&D strategist focused on workplace learning, AI governance, and performance improvement. Connect with him on LinkedIn.