AI Agents Are Not Magic—They Are Executing A Process

Ian Gotts, founder and CEO at Elements.cloud.

It has been almost two years since ChatGPT launched and we started to see the potential of generative AI. At Elements.cloud, we embraced it immediately, and AI has been integrated into every aspect of our core operation—from product management and marketing to business excellence, IT and finance. We’ve also looked at every aspect of our Elements.cloud platform and integrated AI to make our users more productive.

What’s truly remarkable, however, is the cultural transformation it has sparked. Rather than approaching AI with skepticism or hesitation, our teams have embraced it with curiosity and innovation, asking, “Why not?” or “What if we tried this?” We've fostered an environment where experimentation is encouraged, failure is seen as a stepping stone to success and meaningful results take time.

Over the last year, we’ve seen the potential of agents to drive the next wave of AI. But it was a huge engineering effort. Now we are seeing platforms like Salesforce's Agentforce make implementing agents achievable. It is still early days, but we’ve already implemented agents that are performing repetitive, boring, complex tasks. This is freeing up people to do more rewarding, effective work.

First, let’s define what we mean by an AI agent and how it differs from bots or your conversation with ChatGPT. The newest bots can understand natural language but have predefined decision trees that determine their behavior. An AI Agent is provided with data, automated workflows and guardrails that tell it when to pass things off to a human. It can understand natural language requests and then plan which resources to use to deliver an outcome.

The potential of AI agents is still emerging, but when well-designed for a clear use case, they can transform a user experience. For example, an AI agent could handle product returns. By understanding the product, customer and related policies, an agent could process a return, exchange or warranty claim.

That AI agent is not magically becoming a support agent. It is delivering a well-defined process that a human was performing, within a clear scope. The power is that the user is asking questions using their language, and the AI agent is interpreting it and deciding what actions it can take based on what capabilities it has been given.

This means AI agents need to be designed based on the "jobs to be done" (JTBD). The best approach is to draw a UPN process diagram for the AI agent. That will clarify the scope and the expected outcomes. It will also identify the points of handoff back to a human. For building our own AI agents, we’ve realized that you need to think through the process in far more detail than you would if a human was delivering the process. The AI agent doesn’t have the company/contextual common sense. It only knows what you’ve told it about. You need to be far more explicit about the rules and guardrails.

AI can use the process diagram to build the AI agent—writing the instructions and building the actions—which are stored with the diagram. This enables it all to be version-controlled, which is important as there will be a high level of iteration to get the AI agent to perform as expected. This is because building AI agents is effectively “programming using natural language,” with all the potential ambiguity that it introduces. Part of this is because we are still learning how to build AI agents. Over time, there will be well-defined design patterns and prebuilt AI agent templates that make it easier to build high-performing agents quickly.

Some success factors we’ve identified are:

• Go slow to go fast. You will deliver results faster if you have a mature implementation approach, process-led change, data governance and metadata management. Even if your organization does not feel ready to implement an AI agent, then now is the time to start working on these best practices.

• It all starts with the process. Use a UPN process diagram to get agreement on those processes down to a level of detail that enables you to design an AI agent using instructions and actions. Skipping this and reworking and iterating your way to an acceptable AI agent will take longer, knock your confidence and cause you to lose executive support.

• Start small; think big. Building AI agents will be iterative. Focus on narrowly scoped use cases to build experience. Then expand the capabilities of the AI agent as you gain confidence. The learning you get from this first AI agent is as important as the ROI.

• Have a repeatable approach. Building AI agents for the enterprise at scale is like any other tech-led transformation. It requires a proven, repeatable approach. This enables AI agents to be delivered quickly, meet the true business needs and have the required level of governance. Remember, AI agents are your brand ambassadors.

• AI agents can help build AI agents. We’ve designed a repeatable implementation cycle, and this standardization has enabled us to build AI agents that can accelerate the implementation cycle. AI agents can build the first draft, but your expertise is still critical.

• Enable tooling. With every transformation change at scale, you need a platform to help manage this. AI agents need to be planned, designed, built, trained, deployed and monitored. They also need to be integrated into your core systems as they are reusing data and functions.

AI agents could be as disruptive to the delivery of apps and their underlying business models as the shift from on-premise to cloud was over 20 years ago. Those organizations that are best placed are already in good shape with well-understood processes, strong data governance and effective metadata management.

But they need to lean in and explore the art of the possible. No organization can afford to use compliance, data quality or risk as excuses to sit on the sidelines. The future winners have already started; they are piloting, experimenting and learning. And they are accelerating away from the pack—who may never catch up.

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?