I use generative AI for writing, coding, and quick visualisations. I try to use the areas where Generative AI is really good and try to not use where I think I can be faster or better. Recent surveys point that, companies are also rewiring processes and governance to drive bottom line impact, not treating AI as a shiny add on, which is the only way the promised productivity shows up in practice (Singla, Sukharevsky, Yee, Chui, & Hall, 2025).
For text output, the clear advantage for me is structure of the text. I start with an outline generated to a brief, then I rewrite in my own words and verify every citation the Generative AI made. For coding activities, agent based tools help me the most when I ask for a plan first, then apply changes in small, testable steps.
Trough the experiences I have made, trying out different AI tools over a week I even changed the way I work. Now, I write a brief, not a vague prompt with goal, audience, constraints, trusted sources, and what to avoid. I ask for a plan, I approve it, then I let the tool make edits that I can test quickly. I keep a short verification routine, check source links, trace numbers back to a document, run code on a small sample, and log what came from the model and what came from me. It looks slower on paper, in practice it prevents rework and saves time.
For me the takeaway is simple. Generative AI is already useful for getting to a strong first draft in writing, code, and analysis, but only if you design the workflow and the checks around it. Treat the model like a diligent assistant that drafts while you decide, and the gains are real. Treat it like magic, and you will spend your time fixing avoidable mistakes.
References
Singla, A., Sukharevsky, A., Yee, L., Chui, M., & Hall, B. (2025, March 12). The state of AI: How organizations are rewiring to capture value. McKinsey & Company. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
