I sat down with David Irecki, CTO for APJ of Boomi, under the pretext that I felt burned by the Agentic AI hype. At the end of last year, AI boosters were heralding the “Agentic Age” as the next watershed moment in AI, coming quickly on the heels of the progress of generative AI. 2025 has come and mostly gone, and while there are some launches and use-cases that show the potentials of agentic AI, we haven’t seen even a fraction of the impact of AI chatbots.

One of the key differences between generative and agentic AI is that while generative AI for the most part stays within its box (say ChatGPT, CoPilot, or Claude;), AI Agents can be created and designed to work across apps and workflows. The common example is that you can interface with it like a chatbot, typing in a request, and then it can go online, use a browser to do a search and make reservations, and then pop the booked plans into your calendar. This assumes a lot of integration across your apps and accounts, and a lot of trust in letting the agent make decisions on your behalf.

Boomi works more on the business and industry level, supporting mostly MNCs and large enterprises to effectively adopt Agentic AI workflows. I share with David that I’ve used some of the off-the-shelf AI agents, but they haven’t really worked for me.

So David starts to tell me how Boomi approaches this. “If you use a model off the shelf that's pre-trained with some data, you might ask it a question, and it might give you an answer, and you go, Oh, that's not what I was expecting, because it's not related to my business…So that's where context is key…that's what Boomi is great about providing…we provide that interoperability layer. We connect all the systems together for data liquidity. We enhance that data through data quality to enable AI to have that right context.”

We both discuss the much-cited MIT study about failed AI Adoption in companies. Irecki notes that figures from MIT show that 95%of AI projects are failing. While that failure rate might have made some business leaders colder on the idea, we both agree that the study actually shows the proper path forward with AI Adoption. The vision for AI isn’t wrong; it’s that companies lack the right foundations like “clean data, solid integration and good governance” to be able to scale AI.

David is a CTO but the more we discuss, the clearer to me it becomes that the technology is here and it’s ready. I have seen the technology at work. The current roadblock to broader adoption of AI broadly and Agentic AI specifically isn’t a technical problem, it’s a business and strategy problem.

Irecki confirms this view. “You're exactly right,” he says. “You start with integration… you need to eliminate those data silos so that agents have the context to act. And that's not just a technology discussion. Boomi exists to help you do that, but you have to take the step as a business to do that.” He adds that governance is a similar challenge. “The tools for governance around AI exist today…that is also a business discussion around what guardrails do you want to put in from a business perspective, and what ethics and risk?” The technology is there today and is the linchpin for enabling AI, but there is still a lot of discussion around how to have a full end to end successful approach.”

That’s why it’s so important that part of Boomi’s process when working with prospects is a strategy workshop, where they break down workflows and identify where AI Agents could be most beneficial.

For example, Irecki shares a use case in claims management or order-to-cash for a retailer or finance company. “We have an AI agent that's integrated through Boomi into the business that can monitor in real time data across the ERP, the CRM and billing systems,” he explains. “Because it's connected and in real time, what it's able to do is reconcile records, it's able to flag anomalies and potentially trigger approval so you have that human in the loop component automatically, while… maintaining full audit trails”. The agent can even escalate exceptions to humans if its confidence is low. For the customer, this agent can "save us hours per transaction cycle and ensure that those decisions are consistent, compliant and fast.” It is a goal-driven agent embedded in a workflow, not just a simple chatbot.

As we close the interview, I feel that David has helped me rethink my view on Agentic AI. It wasn’t just a buzzy word or promise that didn’t happen, but it’s something that still has the potential to be transformative to businesses. The tech is waiting, but the strategy and bravery of leaders to explore adoption are now the keys.