Image credit: Pexels
From customer support and cybersecurity to marketing and creative services, organizations are discovering both the opportunities and challenges of deploying AI agents at scale.
As agentic AI systems become more capable, business leaders are beginning to rethink the way work gets done. Rather than simply automating isolated tasks, organizations are redesigning workflows around AI agents that can plan, adapt, and execute actions with minimal human intervention. Across industries, early adopters are finding that success depends not only on the technology itself but also on how it is integrated into existing processes and teams.
The Importance of Shared Context
One of the biggest hurdles facing multi-agent systems is maintaining context across workflows. According to Diptamay Sanyal, Principal Engineer at CrowdStrike, AI agents often struggle because they lack the shared understanding that keeps human teams aligned.
“One of the core concepts of a business workflow is you need a shared context across all your agents, because they need to know what you’re building on. With human teams, we do stand-ups and syncs — with agents, you don’t have those. Whatever they hand off, the other one only gets that context. That’s where the quality suffers,” said Sanyal.
The challenge becomes particularly evident in complex business functions such as sales and marketing, where information is constantly evolving and collaboration depends on historical knowledge.
For organizations beginning their AI journey, Sanyal recommends focusing on practical and measurable applications first.
“I would say don’t start with a complex thing like finance. Start with the simpler stuff like customer support — that’s the easiest one, because it’s a proven use case. You have your knowledge base, you have your LLMs, and you can track exactly how effective those agents are,” he said.
Even with proven use cases, adoption takes time. Leaders must give employees room to learn how to work effectively with advanced AI systems.
“You need to have patience. I don’t think it’s going to happen overnight — many times the results are not that great. People also need time to understand how to use these tools, because most folks are still comfortable just going to a chat interface and typing in stuff. But the most advanced AI features don’t come from that,” Sanyal added.
Warning Signs Against the Automation Fallacy
While enthusiasm around agentic AI continues to grow, some experts caution that organizations often deploy the technology before defining the problem they want to solve.
Quentin Reul, Ph.D., Director of Global AI Strategy and Solutions at Expert.ai, argues that businesses frequently focus on the latest innovation rather than the underlying workflow challenge.
“People are looking at the latest technology and trying to retrofit it back to a problem — and that will always fail. The way you solve it is by leveraging frameworks like jobs to be done: look at the problem people are trying to fix first, then see whether the technology is the right one,” said Reul.
His concern extends to governance. In regulated environments, unrestricted AI agents can introduce significant risks, making guardrails essential.
“With agentic AI, if you let an agent do whatever it wants, well, it’s going to do whatever it wants. We’ve heard stories of agents redistributing money from bank accounts, databases being deleted. So you want to put guardrails in place — and in a lot of cases, those guardrails are deterministic. A rule-based system is going to be a lot more effective at controlling what’s going on than an LLM,” he said.
Reul also noted that simply layering automation onto existing processes does not guarantee meaningful improvement.
“People have taken the traditional process of how communication works and put an automation on top — transcription, very useful — but it has not changed the way they’re leveraging the information. You’re probably just sending another meeting to get to the person and ask them the right question. We’ve automated but actually added more friction,” he said.
Keeping Humans in the Loop
For digital marketing firm e-intelligence, agentic AI is being used to remove repetitive work while preserving human oversight.
“We are creating agents for marketing automation, testing multiple workflows, trial and testing. We cannot say that we are experts of agentic AI, but yes, we are testing and trying these things for repeated marketing tasks,” said Miral Garala, Assistant Delivery Manager at e-intelligence.
One example is the company’s Monthly Report Automation agent.
“We have created an agent called Monthly Report Automation. It checks all of our marketing platforms — Google Analytics, SEMrush, SCRanking — and creates a detailed report for us. We get a report within a few minutes, and then our expert team members verify it before sharing with our clients,” Garala explained.
That verification step remains critical.
“We are using AI, but we are verifying from our side before taking any actions on the AI responses. In the current situation, human verification is needed — AI models are trained for particular algorithms, so we cannot 100% rely on the AI,” she said.
Focusing on Time Savings
For Joseph Suarez, founder of Hollywood Method, the impact of agentic AI is measured less by direct revenue and more by efficiency gains.
“AI isn’t just about can it make you money — it can save you time. Time is worth more than any money you think you’re making right now, because more time means more time to learn more things. More things you’ve learned means more you can charge to make more money,” Suarez said.
After restructuring his agency around AI-powered workflows, he reported significant operational changes.
“My agency went from a team of 14 to two — and I’m not adding to unemployment more than I already have. I gave raises instead of always needing to hire more people. Efficiency is up, costs are down,” he said.
For Suarez, the broader shift is about redefining human roles within organizations.
“The monkey work that you were paid to do should have never been your job. Embrace the creativity. Embrace the fact that you are now an operator and not just the machine,” he added.
Leaders Define the Future of Agentic AI
Across cybersecurity, enterprise AI, marketing, and creative services, a common pattern is emerging. Organizations seeing the greatest benefits from agentic AI are not treating it as a replacement for human decision-making. Instead, they are using it to handle repetitive, complex, and time-consuming work while maintaining strategic oversight.
As businesses continue to experiment with AI agents, the defining factor may be neither the sophistication of the tools nor the speed of adoption. Rather, it is the ability of leaders and teams to understand where automation adds value and where human judgment remains essential.
