Agentic AI: Use Case That Drive Revenue

2025 marks a turning point in AI adoption. This is not going to be simply about smarter algorithms, but the next step in autonomous, agentic systems capable of reasoning, acting, and collaborating with independence. 

According to our 2025 market research on data & AI, nearly half of organizations use or are developing generative AI assistants, and, for many, AI agents are a strategic priority. But which use cases are driving revenue? And how to replicate successful companies? In this post, we’ll take a deep dive based on the results from our Data & AI Monitor!

What Makes AI "Agentic"?

First, some definitions. What’s an agent? There are three key traits that set them apart:

How can you succeed when implementing agents? Naïvely, you would think that tech plays a big role: at the end of the day, these agents are complex software systems that have to react and deal with a chaotic reality.

While that might still play a role, our Data & AI Monitor shows that organizations that understand where and how to apply AI and couple it with a strong strategy are more likely to succeed.

High-impact agentic AI use cases

Financial Compliance

Our report highlights how financial services are leading the way in AI agent adoption, with 40% of companies having adopted them.

And we could have expected it: strong governance, high-quality data, and the large impact automation and AI can have on financial processes are prime factors of speedier adoption of AI in financial services.

According to research from McKinsey & Company, financial crime is a banking sector where the rewards for agentic AI are greater. Know-your-Customer (KYC) modules and Anti-Money-Laundering (AML) activities require many manual, repetitive tasks, sifting through large quantities of data and disparate systems. Agents, with their autonomous decision-making, contextual adaptability, and multi-step execution, can automate much of the work, leaving only QA and escalation paths to financial crime specialists.

Our report, unsurprisingly, highlights that companies need to trust AI systems to reap the benefits in such sensitive areas—the memory of a three-quarters-billion-dollar fine for a European bank failing to implement KYC processes is still vivid — and prioritizing explainability of and logging the agent’s decisions remains a priority for adoption.

KYC and AML are not the only high-impact agentic AI cases in finance. Recently, a multinational bank deployed an agentic AI solution that increased customer satisfaction and reduced deflection to the customer support department by as much as 20%. As with financial crime, understanding the agent’s behavior turned out to be a critical factor for success.

Autonomous Customer Service Resolution

Customer support employees spend most of their time on repetitive, basic issues like password resets or order tracking. No wonder chatbots have been deployed for years to automate some of it. However, whenever a customer went beyond the rigid scripts, the chat had to be routed to a person, only to reset the password to “Welcome97” (the previous one, in fact, was “Welcome96”, so it couldn’t be reused. Security, go figure!) and wasting everybody’s time while doing so.

It’s no wonder that Large Language Models, with their superior ability to figure out what we mean, are being employed left and right with customer service agents. These agents are given tools to access customers’ purchases, past tickets, knowledge bases, and the infamous password reset capability — serving us autonomously for much longer before rerouting us to a person.

I experienced their ability firsthand the other day. Almost two years ago, I bought a wired Ring doorbell. Just as the warranty was about to expire, the bell became completely unresponsive. Ring.com was kind enough to send me a new one, but they sent me two emails with two different shipping labels to send the old one back. I shrugged, picked one, printed it, and shipped it.

Two weeks later, I received an email from Ring.com saying that they still hadn’t received my defective bell (I thought that the unused shipping label was triggering some alerts). So, I hit up the website, opened the chat, and there was an AI agent waiting for me. And I knew it was an agent because it had tool calling and autonomous decision making. In fact, after explaining the situation, the agent told me I didn’t have to worry, and it’d make a note of the situation in my file.

Two weeks passed, and boom, another email from Ring.com. Again, they still hadn’t received the bell. This time, the chat brought me to a helpful person who read the note written by the agent and reassured me that the emails would eventually stop. From this I gathered that, while the agent took the decision and wrote the note, it didn’t seem to have access to the alert system triggering those emails. Perhaps something to add in the next iteration?

All in all, I was really impressed by the agent’s abilities, but Ring.com is in the minority: our report highlights that only 35% of successful AI implementations integrate with core business processes, such as CRM and helpdesk systems. Even if in the minority, however, they are not the only ones. One of our case studies highlights how Agentic AI helped a global cruise operator boost operational efficiency and reduce missed engagement opportunities by 80%.

Why Most Agentic AI Implementations Fail

Despite these successes, the 2025 Data & AI Monitor reveals the harsh truth about why Agentic AI fails: organizations need guidance on where and how to apply AI. In fact, more than 50% of the respondents gave a neutral or negative reply to this question. 

So, if they don’t know where and how to apply AI, what are they building? There are two possible paths: they’re not doing anything, or they are building solutions looking for problems. This is a common trait of organizations that don’t put the business in the driver's seat but are pushing AI from a tech angle.

We see these themes throughout the survey. The key organizational elements for AI success are stakeholder management and storytelling, which are closely related to putting the business front and center when building AI systems.

On top of it, we observe few organizations have a clear AI strategy: the rest lack clear planning, strong governance, workforce readiness, and disciplined measurement.

The latter is especially troublesome: only 33% of companies say they are tracking AI's actual business value which was shocking to me and probably one of the main culprit companies cannot move beyond a pilot or proof-of-concept: scaling an AI system can be costly — especially when incorporating it within operational processes — and without a clear value, there is no business case to do so. 

How to succeed with Agentic AI

How to succeed is then simple but requires a lot of work and the recipe is in the previous section:

The technology side brings some ingredients to the recipe as well. The top elements are:

That’s not all, of course. Especially when dealing with agents, well-set decision boundaries — defining how far they can take their decisions and what they can do on their own — and definition of when human intervention is required are a prerequisite for successful agent deployments.

The agentic AI revenue opportunity

Not all organizations are seizing the opportunities agentic AI offers them, due to all challenges listed above. Although the fix is simple, these companies lack the drive and guidance to execute a well thought-out strategy. And that is no surprise. A strategy lays out the investment but also the expected returns. And since the majority of companies have no clue where and how to apply AI, they don’t know what to expect when it comes to ROI.

An easy fix is therefore to focus on proven revenue drivers (sales, supply chain, product) while avoiding common pitfalls—as outlined here and in our Data & AI Monitor — to turn agentic AI from a buzzword into a well-oiled profit engine.

Experimenting and piloting can be a great way to taste the wine before buying the bottle, but success lies one step beyond, when companies operationalize their AI systems at scale, embedding them in the organization and aligning them on its business goals. Strong leadership will be a guidance in overcoming these issues and reaching true AI adoption.