- Agentic AI systems consume up to 136.5 times more energy per query than traditional chatbots, according to a KAIST study, raising sustainability concerns.
- In manufacturing, agentic AI enables real-time decision-making across inventory, quality, and workforce management, but requires edge computing and data integration.
- Gartner warns that HR leaders must prepare for autonomous workflows, while AWS and partners are building tools to measure business value and scale agentic solutions.
- Data governance and cultural change are critical: without clean, unified data and a shift from control to orchestration, agentic AI risks remaining a proof-of-concept.
The conversation around artificial intelligence has shifted from simple chatbots to something far more ambitious: agentic AI systems that can plan, reason, and execute multi-step tasks autonomously. Unlike traditional generative models that respond to a single prompt, these agents break down complex goals, call external tools, and adapt their actions in real time. But as this technology moves from labs into factories, offices, and cloud platforms, a series of challenges are emerging that go well beyond the usual hype.
Recent studies and industry announcements paint a picture of both promise and pressure. From energy consumption that could rival national grids to the need for new leadership mindsets, the path to widespread agentic AI is paved with hard trade-offs. Here’s what the latest research and expert insights reveal about where this technology is headed and what it will take to make it work.
The Hidden Energy Cost of Autonomous Agents
A study from the Korea Advanced Institute of Science and Technology (KAIST) has put a spotlight on a critical issue: agentic AI can consume up to 136.5 times more energy per query than a conventional chatbot. The reason lies in the architecture. While a chatbot generates a single response, an agent may invoke a large language model multiple times, wait for external tool responses, and coordinate several steps before delivering a final answer. In one scenario, a 70-billion-parameter model used an average of 348.41 watt-hours per query.
The problem is compounded by idle time. According to KAIST, agents can take up to 153.7 times longer than chatbots to complete a task, and during that period, GPUs can remain inactive up to 54.5% of the time while waiting for external data. Yet those GPUs are still powered and reserved, adding to the electricity bill without doing productive work. The institute’s worst-case projection—13.7 billion daily agent queries, equivalent to Google’s search traffic—would require 198.9 GW of data center power, roughly half of the average U.S. electricity consumption.
This isn’t just a theoretical exercise. The International Energy Agency estimates data centers consumed 415 TWh in 2024, with a potential doubling to 945 TWh by 2030. In the U.S., the Department of Energy projects data center electricity use could reach between 6.7% and 12% of total national consumption by 2028. Agentic AI accelerates this demand at a pace that electrical grids struggle to match, given the long lead times for new generation and transmission infrastructure.
Transforming Manufacturing with Autonomous Decision-Making
In the industrial world, agentic AI is being hailed as the next leap beyond automation. Zebra Technologies, a leader in enterprise asset intelligence, sees it as a way to turn factories into self-organizing entities. Instead of following pre-programmed rules, agents can interpret real-time data from inventory, production, sales, and logistics to make coordinated decisions without human intervention.
Zebra identifies three areas where the impact will be most immediate:
- Intelligent inventory management: Agents analyze consumption trends, supplier availability, and seasonality to optimize stock levels, reducing waste and capital tied up in goods.
- Quality assurance: Computer vision powered by AI inspects every product for defects, mislabels, or illegible dates, catching issues without stopping the line.
- Workforce optimization: In large plants, agents automatically adjust shift schedules, assign tasks based on skills, and cover absences, freeing supervisors for strategic work.
To make this work in real time, edge AI is essential. New industrial devices with neural processing units (NPUs) run models locally, cutting latency and ensuring operations continue even if cloud connectivity drops. Early results are promising: a partner case study showed a 98% faster incident diagnosis and a 50% reduction in operational variability, along with at least a 50% drop in emergency transport costs.
Redefining Work: The CHRO’s Role in an Agentic Era
Gartner’s latest analysis warns that the rise of agentic AI will force human resources leaders to rethink their entire approach. Unlike generative AI, which assists, agentic AI executes entire workflows autonomously, from initial recruitment screening to onboarding and payroll management. This shifts the CHRO’s job from overseeing technology adoption to managing cultural change and redesigning job roles.
The key challenge is governance. Autonomous agents making decisions about hiring, performance, or compensation must be audited for bias and privacy compliance. HR leaders need to establish clear frameworks for when and how agents can act, while also identifying which human skills remain critical in a world where routine tasks are handled by software. Gartner emphasizes that this is not a technology problem but a strategic one, requiring active leadership from the top.
Building the Ecosystem: AWS and Partner Initiatives
Amazon Web Services is betting big on agentic AI as a revenue driver for its partner network. At the AWS New York Partner Summit, the company launched several initiatives to help partners demonstrate business value. The new Business Value Realization (BVR) program provides methodologies, benchmarks, and outcome-based funding to prove ROI, addressing a McKinsey finding that only 39% of organizations can measure AI’s impact.
AWS also introduced a suite of agentic tools—including Amazon Quick, Kiro, and agents for security, DevOps, and FinOps—with free trial periods. Marketplace fees for professional services were cut from 2.5% to 0.5%, and new AI-powered sales features help partners identify opportunities. The company estimates that agentic AI could generate up to $200 billion in new value for the partner ecosystem over the next few years.
From Experiment to Operation: Integrating Agents into Business Processes
For many organizations, the real hurdle isn’t the technology itself but how to embed it into existing workflows. Carlos de Marco, VP of Product at Auronix, argues that the true value of agentic AI lies in its ability to connect fragmented processes. In retail, for example, an agent can guide a customer through a purchase, check inventory, coordinate payment, and initiate support—but only if it has access to CRM, payment systems, and logistics data.
This requires a shift in mindset. Instead of treating AI as a standalone tool, companies must see it as an operational layer. The most successful organizations are those that ask not ‘what can AI do?’ but ‘which process do we want to transform?’ In Mexico, e-commerce reached 941 billion pesos in 2025, yet many companies still operate with siloed data, limiting the impact of any AI deployment.
Data, Governance, and Talent: The Foundations for Success
SoftServe, a global digital engineering firm, highlights that agentic AI’s effectiveness depends entirely on the quality of underlying data. A joint study with MIT Technology Review found that 72% of organizations expect AI agents to manage most or all of the software lifecycle within two years, but in Latin America, data fragmentation remains a major barrier. 58% of business leaders say key decisions are based on inaccurate or inconsistent data.
John Howard, SoftServe’s VP for Latin America, stresses that an agent is only as reliable as the data and permissions behind it. Companies must first unify their data architecture, establish governance, and move from siloed pilots to a shared AI infrastructure. Talent is another piece: the demand is shifting from traditional developers to specialists in automation, data, and agentic engineering. SoftServe itself trained 80% of its staff in AI during 2025 through its AI Edu Space platform.
Finally, the cultural shift cannot be ignored. Leaders must evolve from controlling tasks to orchestrating workflows where humans and agents collaborate. The biggest blind spot is treating AI as a tool to be supervised rather than as a management layer. Only by designing shared workflows and critical human oversight can companies turn agentic AI from a promising experiment into a sustainable competitive advantage.
Agentic AI is not just a faster chatbot or a smarter assistant. It represents a fundamental change in how work gets done—from energy-hungry data centers to self-organizing factories, from automated HR processes to integrated customer journeys. The technology is advancing rapidly, but its success will depend on solving the hard problems of infrastructure, data quality, governance, and leadership. The next big race in AI may not be about building smarter models, but about making them work without breaking the systems they run on.
