AI agents are quickly becoming one of the most powerful tools in the enterprise AI toolkit. According to Cloudera’s latest Agentic AI Survey, 96% of enterprises plan to expand their use of AI agents. But what’s driving this surge, and how can organizations turn hype into results?
Abhas Ricky, Chief Strategy Officer at Cloudera, joined The AI Forecast to discuss the real momentum behind agentic AI.
Here are some key takeaways from that conversation.
AI agents are no longer a theory. They’re digital coworkers.
Paul: Let’s talk about the big headline from the most recent report on agentic AI, citing that 96% of enterprises plan to expand their use of AI agents in 2025.
Abhas: If you look at the last 40 years of technology, the dream has always been to simplify complex tasks. Thanks to agentic AI, we’re seeing the evolution of that dream come to life. These aren’t just chatbots or assistants. They’re digital workers with memory and autonomy. They process tasks independently, adapt based on past behavior, and don’t need to be constantly updated or instructed.
That makes them so powerful, and why enterprises already see 10 to 100x productivity improvements.
The biggest blocker? It’s not tech. It’s trust.
Paul: On the other side of that 96%, what’s slowing workflows down? What’s keeping that other 4% on the sidelines?
Abhas: That 4% points to the core blocker of enterprise adoption: data privacy. Over 53% of enterprises said scaling AI is their biggest challenge.
Agents operate across multiple systems, such as CRM and sensitive medical or financial data. Without clear boundaries, you’re introducing risk at every step. It’s not just about protecting the data; it’s about ensuring agents know who can access what at what time. That means robust policies, metadata governance, APIs, and enterprise-grade authorization frameworks.
Start with fast wins, then scale.
Paul: Let’s talk about implementation. Where should enterprises start when scaling agentic AI to see value fast?
Abhas: There are always cases where low-hanging fruit can deliver quick wins. That’s why we work with customers to break down large, complex initiatives into manageable micro-workflows that show value fast.
For example, a global bank we partnered with wanted to reduce its mortgage processing time from four weeks to just six hours. It’s a bold goal, but we didn’t try to overhaul everything at once. We started by automating sub-workflows like tax document review and data benchmarking. That’s where the early ROI appeared, and from there, we expanded.
The key is to break the use case into smaller workflows. Some will be easy and fast to implement, others more complex and time-consuming. Our recommendation is always the same: prove value early, pick quick wins first, earn the right to go deeper, and keep up the momentum. And remember: Wherever possible, bring the model to the data, not the data to the model—most enterprise data still remain locked where it was created.
The road ahead involves full autonomy and talent arbitrage.
Paul: In the next 12 to 18 months, what specific advancements do you expect in autonomous agents?
Abhas: You’ll see companies offering fully autonomous agents, especially as new protocols like MCP and ATA make systems more interoperable. But there’s a bigger shift happening, too, where AI is becoming geopolitical. Countries and companies are building sovereign AI stacks, and we’ll see more demand for full-stack LLM engineers who can build enterprise AI from the ground up.
Certain jobs will now be augmented. But the number of jobs created will outnumber the reduction in jobs because of AI. The nature of the job will just be different. I’m optimistic that we will never lose the value of human input – regardless of how advanced the LLMs and systems become.
Catch the full conversation with Abhas Ricky on The AI Forecast on Spotify , Apple Podcasts, and YouTube.
