GIỚI THIỆU VỀ TẬP NÀY
Building AI agents for production means choosing how much autonomy to give them — and that answer keeps changing. Aakriti Bhargava, VP of product engineering and AI at Revionics, has spent 20+ years in applied AI. When generative AI arrived, the hardest part wasn't building features. It was a security conversation she didn't see coming, and an architecture debate that's still evolving.
This conversation covers how Revionics thinks about the balance between deterministic and autonomous agent behavior, why they chose not to fine-tune foundation models and focused engineering effort on the system around them instead, and what happens to engineering value when code generation becomes trivial. Aakriti also shares how generative AI compressed RFP timelines from weeks to days, why code review is the new bottleneck, and her candid take on AI coding tools making junior engineers worse at problem solving.
This is the third and final episode in a three-part series on building the infrastructure foundation that makes everything else possible. The first two episodes cover Clari's petabyte-scale migration and Revionics' cloud migration that finished a year ahead of schedule.
Get Insight's modern infrastructure solutions because you'll see how enterprises are building AI-ready foundations and shipping agents to production: https://www.insight.com/en_US/what-we-do/expertise/modern-infrastructure.html
Subscribe and follow Insight On for new episodes every week.
#AIagents #GenerativeAI #ProductEngineering #EnterpriseAI #InsightOn
Chapters (5–12)
-
00:00 — Welcome and introduction
-
01:28 — What Revionics does and its AI history
-
02:01 — Why generative AI integration was harder than expected
-
04:04 — What generative AI makes possible that wasn't before
-
05:19 — Buy vs. build for generative AI and pricing AI
-
07:07 — Deterministic vs. autonomous agent architecture
-
10:00 — Why every leader needs to understand agent decisions
-
11:10 — How AI coding tools changed engineering productivity
-
13:17 — Why junior engineers may be hurt by AI tools
-
15:35 — Rebuilding customer trust for generative AI
-
17:19 — What engineers need to unlearn in the AI era
