The framing of "kill vs. not kill" misses the more interesting structural shift: AI agents don't eliminate data engineering they eliminate the routine execution layer and concentrate the remaining value at the top. What I'm seeing in practice is one principal architect with three to five well-configured agents running work streams that previously needed a team of four. The on-ramp problem is real though junior engineers used to learn the platform by doing exactly the tasks agents now absorb. I wrote about this operating model and the governance layer that makes it actually hold in production: https://www.nilayshah.me/ai-agents-are-shrinking-middle-why-they-make-1-expert-more-powerful-than-a-full-team/
Thanks for sharing my article Daniel! 🙏
The framing of "kill vs. not kill" misses the more interesting structural shift: AI agents don't eliminate data engineering they eliminate the routine execution layer and concentrate the remaining value at the top. What I'm seeing in practice is one principal architect with three to five well-configured agents running work streams that previously needed a team of four. The on-ramp problem is real though junior engineers used to learn the platform by doing exactly the tasks agents now absorb. I wrote about this operating model and the governance layer that makes it actually hold in production: https://www.nilayshah.me/ai-agents-are-shrinking-middle-why-they-make-1-expert-more-powerful-than-a-full-team/