← Multimodal Sources 🕐 3 min read
Multimodal Sources

From Models to Agentic Systems: The Next Enterprise AI S-Curve | Andrew Ng joins VB Transform 2025

> I came out and started using the word 'agentic AI' and proposed to the community, let's stop having this debate.

Show: Beyond the Pilot · Publisher: VentureBeat · Host: VentureBeat editorial

Episode URL: https://www.youtube.com/watch?v=lB8jythC-vQ

Publish date: 2026-04-13
Duration: 1189.0s
Default source credibility: HIGH — Named F500 practitioners on-record with production metrics. VentureBeat editorial vetting. Treat vendor-sponsored segments as MEDIUM.

  • The term “agentic AI” was coined to shift focus from defining “agents” to building iterative, multi-step AI workflows with varying degrees of autonomy.
  • Agentic systems, even with less advanced models, can outperform more powerful models on complex tasks by breaking them down into a sequence of steps.
  • The primary enterprise opportunity is automating complex, multi-step business workflows, representing the next major AI adoption S-curve.
  • The cost and time to prototype AI applications have plummeted, enabling rapid experimentation. Enterprises should create sandboxes for safe, fast iteration.
  • While reasoning is being baked into foundation models, agentic workflows remain essential for custom, boutique enterprise tasks that require specific processes.

Extracted quotes

# Credibility Speaker Org Timestamp Topic Quote
1 HIGH Andrew Ng (Founder) DEEPLEARNING.AI 01:41 12-agent-workers I came out and started using the word ‘agentic AI’ and proposed to the community, let’s stop having this debate. Let’s think of it as having degrees of autonomy, where some things are less agentic, and then there are things that can be highly autonomous and carry out a complex sequence of tasks all by itself without human intervention.
2 HIGH Andrew Ng (Founder) DEEPLEARNING.AI 16:23 07-adoption-challenges For many businesses, if we can create sandboxes that let innovative teams iterate really quickly and build prototypes—because it’s running on a sandbox with limited private information, not shipping outside, not using the company brand, not risking damaging the brand—that lets teams run really quickly. Then we can see what works and only then invest in those things to make sure it’s responsible.
3 HIGH Andrew Ng (Founder) DEEPLEARNING.AI 15:02 02-corporate-tools There are so many projects that previously would have taken me three months and six engineers to build, that today, I or one of my engineers will build literally on a weekend. This lets us prototype really fast and decide what to take to production.

Per-quote detail

1. Andrew Ng — DEEPLEARNING.AI (01:41)

I came out and started using the word ‘agentic AI’ and proposed to the community, let’s stop having this debate. Let’s think of it as having degrees of autonomy, where some things are less agentic, and then there are things that can be highly autonomous and carry out a complex sequence of tasks all by itself without human intervention.

  • Credibility: HIGH — The speaker is a leading AI practitioner providing the origin story and definition for a key industry term.
  • Topic tag: 12-agent-workers

2. Andrew Ng — DEEPLEARNING.AI (16:23)

For many businesses, if we can create sandboxes that let innovative teams iterate really quickly and build prototypes—because it’s running on a sandbox with limited private information, not shipping outside, not using the company brand, not risking damaging the brand—that lets teams run really quickly. Then we can see what works and only then invest in those things to make sure it’s responsible.

  • Credibility: HIGH — The speaker, a top AI practitioner, provides a specific, actionable strategy for managing AI innovation and risk within an enterprise.
  • Topic tag: 07-adoption-challenges

3. Andrew Ng — DEEPLEARNING.AI (15:02)

There are so many projects that previously would have taken me three months and six engineers to build, that today, I or one of my engineers will build literally on a weekend. This lets us prototype really fast and decide what to take to production.

  • Credibility: HIGH — The speaker is a top AI practitioner providing a specific, unscripted comparison of development timelines before and after current AI tools.
  • Topic tag: 02-corporate-tools

Extracted 2026-04-13T23:39:36 via scripts/podcast_mine.py (Gemini gemini-2.5-pro).