Agentic AI reality check: 5 blind spots for organizations

1. Autonomy's hidden costs beyond subscriptions

Let's address the elephant in the room: cost. This isn't just about SaaS subscriptions or upfront implementation fees. It's about the ongoing operational expenses of running agentic systems.

If we want to move beyond intelligent automation and truly move to autonomous agents, these systems will need to be data hungry, constantly querying, processing and analyzing information. And that means a massive increase in compute, storage and networking bandwidth. Are we ready for that infrastructure overhaul?

And what about the cost of human oversight? Autonomous does not mean hands-off. We'll still need qualified people to monitor, tune and intervene when agents inevitably run off the rails. This is a new kind of labor, a new kind of infrastructure and a new kind of cost. And frankly, most organizations are grossly underestimating it.

This is why the above-mentioned 89% of organizations allocating more budget to integrating, accessing and analyzing their data really need to think about where that money is going. We need to allocate more budget overall for the entire ecosystem of agentic AI. Or, at the very least, recognize that whatever your current budget is, it is not enough. Hidden costs, as always, will rear their ugly heads.

2. The ethical minefield of autonomy and accountability

Agentic AI raises serious ethical issues that we are just beginning to uncover. Who is to blame when an autonomous agent makes a biased decision? Or when it breaks a privacy law? Or when it just gets something wrong?

This lack of accountability is a ticking time bomb. We need the right ethical frameworks, governance policies and audit trails in place before we release agents into production. This is not just about compliance; it's about trust.

And trust, in the age of AI, is nonnegotiable. We can train employees, stakeholders and other partners on the coming challenges and issues, but that alone still might not be enough.

3. Amplifying 'garbage in, garbage out'

We've all heard the saying "garbage in, garbage out." Agentic AI is no different. These systems are only as good as the data they're trained on or the data they use to inform decisions.

And let's be honest: Most organizations are still grappling with data quality issues. In fact, 84% say that it's their most significant challenge when it comes to analytics.

Now imagine those issues amplified by autonomous agents. The potential for biased, inaccurate and downright dangerous insights and decisions is massive.

I realize that it is a tedious topic, but we need to double down on the processes for data cleansing, validation and governance. We also have to put guardrails in place to prevent agents from exacerbating existing biases or creating feedback loops of misinformation. I have yet to see a comprehensive solution that will truly address these problems.

4. The shift from data analysts to agent wranglers

Agentic AI will completely transform the function of data analysts. They won't just crunch numbers and build dashboards anymore; they'll design, train and manage autonomous agents.

This demands a new mindset and skill set -- AI literacy, agentic design, ethical reasoning -- mixed with existing critical thinking skills. Frankly, the current workforce is very much unprepared to do this.

Not only that, but many don't even want to take on that role. We need to invest in extensive reskilling and upskilling of the future workforce, and we need to rethink our recruiting strategies to attract and retain the "agent wranglers" of the future.

5. Agentic AI as a catalyst for organizational collaboration

Business leaders have talked for a long time about breaking down silos. Could agentic AI help finally do it?

With a unified view of an organization's data, agents could do more than intelligently automate workflows within departments. They could span business units to deliver the next level of collaboration.

But this requires a different mindset. We need to shift from data hoarding in business units to cultivating a culture of data sharing and collaboration. Doing so will enable these agents to act as "data diplomats," better connecting technical teams and business stakeholders.

What's next?

So where do we go from here?

We need to start having some of these difficult conversations now, regardless of what vendor hype might lead you to believe. The agents that might one day provide a new and effective way of thinking about your business are already here, but success will depend on practicality, ethics and organizational trust.

We need to build a future in which agents are not just hype-focused, but are good, ethical, socially responsible and accountable. Agentic AI is the way of the future in analytics, so it is up to us to shape that future through negotiation and collaboration in ways that benefit all.

Mike Leone is practice director for data management, analytics and AI at Enterprise Strategy Group, now part of Omdia.

Enterprise Strategy Group is part of Omdia. Its analysts have business relationships with technology vendors.