Rise of Agentic AI
Sachin Chheda
Driving the next generation of IT, Cloud, and AI services and innovation
I had a chance to attend an event early this month on the role of AI in the enterprise. Like it is with many of these events, the sessions featured experts including founders who have incorporated AI at the core of their offerings. Their products represent an evolution of how AI is being used, moving from a copilot to an agent.?
Let's start by understanding these two words. Copilot is AI-powered assistance. It enables individuals and other AI functions to draw on its expertise. Examples include assisting with coding (suggestions & generation, reviews and more), marketing & literary (most popular use case is content generation), and analytics & reporting (predictive analytics, trend identification and more). Agentic AI is a complete use case where AI can perceive and interact with people and other AI functions. They can learn and evolve over time to enhance their performance based on their interactions. Agentic AI involves AI models working through workflows for interactions, serving functions like customer service, gaming, virtual assistants, and more.?
One of the questions that came up a few times across the different sessions was how can an organization take advantage of the AI innovation.?
The most common answer from the panelists was to take advantage of an AI powered service or function like customer service agents that automated support interaction with customers. One of the panelists talked about fine tuning and customizing open source models for their customers to deliver a curated copilot function for software developers. Another talked about creating custom support agents trained in an organization’s knowledge base and support cases, automating customer interactions.?
Almost all of the panelists, both practitioners and founders who have had experience with implementing AI in their operations shared that organizations will need to invest in some customization, including fine tuning/RAG and bespoke workflows for their needs. While most models (LLM etc) out of the box provide some value, investing in customization is a must to get the right return on investment. One of the practitioners who is responsible for their company’s customer experience function had solid advice that organizations should use the opportunity to run an extensive proof of concept (POC) to field test their assumptions including outcomes and investments. They walked through their own experience of a paid POC that helped them identify opportunities to refine their own strategy including how to handle data and the amount of customization they should expect to get reasonable results.?
领英推荐
Another panelist in the following session stressed the importance of investing in training and if needed consulting resources. Their point was that AI initiatives should be treated the same way one should approach other technology investments. Having the right people with the right technology and application skills is important. Since the concept of AI in the enterprise is new, it is tough to find folks with the right talent. But having the partner (ISV, consulting, system integration, managed services, operations) can reduce the burden. I always recommend decision makers to adopt a framework similar to what is presented by Niel Nickolaisen in an article from issue 1.0 of the NEXT magazine, titled 'Decision Framework for Purpose Alignment'. The article discusses decision-making for the optimal use of IT resources, encouraging outsourcing IT apps and services that are not high on the criticality scale.
On other news, Nutanix announced the Enterprise AI offering that now not only supports running on customer’s premises (i.e. private cloud) and at service providers (private and shared clouds), but also supports CNCF compatibility K8s environments including the hyperscalers. Read more here and here (blog).
Here are some more resources on Agentic AI (good reading for IT execs & decision makers).?
Driving the next generation of IT, Cloud, and AI services and innovation
2 周Mike Barmonde -- it was great catching up with you and thanks for sharing your insights on the Agentic AI including shared services, guardrails, embedding data, etc and the details on the Nutanix Enterprise AI creating secure endpoints using RBACs.
CXO Advisor | ex-Persistent, Zinnov, Microsoft, Monitor Group
4 个月Great summary Sachin. The point around “decision-making for the optimal use of IT resources, encouraging outsourcing IT apps and services that are not high on the criticality scale” lacks understanding & companies are not able to take full advantage of the ecosystem because of the perceived notion of keeping things “in-house” in the AI world.