Why Agentic AI Matters for Business

What Are AI Agents, and Why Are They Powerful?

AI agents are autonomous software systems powered by large language models (LLMs) that can reason, plan, act, and learn, often across multiple steps, tools, or data sources. Unlike traditional LLM-based chatbots that answer questions and wait for the next prompt, AI agents do things. They interpret goals, break them down into tasks, access external tools or APIs, coordinate actions, and refine outputs, all with minimal human input.

Key Capabilities:

• Contextual Understanding: Agents remember past interactions and evolve strategies accordingly.
• Tool Use: They connect to APIs, CRMs, databases, spreadsheets, robotic systems, and more.
• Autonomy: They make independent decisions and execute complex, multistep workflows.
• Collaboration: In multi-agent systems, they delegate, validate, and orchestrate with one another for even greater power.

“Agents don’t just interact. They reason, adapt, collaborate—and act on your behalf.” – Deloitte

How Agentic AI Works

An AI agent architecture generally includes:

1. A reasoning model (e.g., GPT, Gemini, Claude)
2. Memory (short- and long-term)
3. Tool access (e.g., APIs, code execution, search)
4. An orchestration layer that plans, loops, and evaluates actions toward a goal

Agents often follow frameworks like ReAct, Chain-of-Thought, or Tree-of-Thoughts to iterate through decisions and dynamically respond to feedback or updated information.

Why It Matters Now

According to UiPath, over 90% of U.S. IT leaders believe AI agents can improve business processes, and over half plan to deploy them in the next 12 months. Businesses using AI agents are already reporting:

• Massive ROI gains from automating high-cognition tasks
• Faster customer service resolutions (e.g., Wiley achieved 40%+ improvement)
• Time savings (e.g., Oracle saved 66 minutes per day per clinician)
• Smarter workflow orchestration, reducing labor-intensive tasks like data synthesis, report generation, and multi-system coordination

Agentic automation, unlike RPA or traditional automation, adapts to changes, learns from patterns, and operates in real time with human oversight when necessary.

Agentic AI Is the Future

This isn’t just another wave of automation—it’s a reinvention of how digital work gets done. The move toward agentic ecosystems will redefine roles, organizational structures, and strategy.

Predicted Future Trends:

1. Agents will become teammates — coordinating across departments and systems.
2. Agent swarms (multi-agent systems) will outperform single-agent tools for creative, analytical, or operational work.
3. Memory and self-reflection will drive personalization in healthcare, finance, education, and customer support.
4. Agentic UX design will become a discipline of its own, merging design, AI, and human-centered orchestration.
5. Every job will have a digital twin—a persistent agent that learns and grows with the worker, automating more each month.

“In 2025 and beyond, agents won’t just power chatbots—they’ll run your workflows, write your reports, update your records, and talk to your APIs.” — RASA

But Tom, what about mother GAIA and the water, noise and the baby unicorns affected by inefficient data centers?

These companies are at the forefront of developing solutions to enhance the energy efficiency of AI systems, addressing both computational demands and environmental concerns. These companies are actively addressing the computational and energy challenges posed by artificial intelligence (AI) through advancements in both hardware and software. Here's an overview of key players and their initiatives:

NVIDIA

NVIDIA has significantly improved the energy efficiency of its AI systems. Over the past decade, their GPUs have achieved a 2,000-fold reduction in energy consumption for training tasks and a 100,000-fold reduction for inference tasks. Network World

Their Grace Hopper Superchip, combining CPU and GPU capabilities, has demonstrated a 4x reduction in energy consumption and a 7x decrease in processing time for financial services workloads compared to CPU-only systems. NVIDIA Blog

NVIDIA's accelerated computing platform, which includes GPUs, CPUs, and DPUs, is designed to maximize energy efficiency across various industries. Energy Central+3NVIDIA Blog+3NVIDIA Blog+3

IBM

IBM is focusing on developing energy-efficient AI hardware and models. Their NorthPole chip, inspired by brain architecture, offers 25 times greater energy efficiency and 22 times faster performance than traditional GPUs for specific AI tasks. Macro 4IEEE Spectrum+2IBM Research+2Wikipedia+2

IBM's analog AI chip for deep learning inference showcases a scalable mixed-signal architecture, aiming to reduce power consumption during AI operations. IBM Research+1EE Times+1

Additionally, IBM is committed to creating smaller, more effective AI models to minimize energy usage, enhancing the sustainability of AI applications. IBM - United States

AMD

AMD has set an ambitious goal to increase the energy efficiency of its processors and accelerators used in AI training and high-performance computing by 30 times over a five-year period. AllianceBernstein

Huawei

In response to export restrictions affecting NVIDIA's H20 chip, Huawei introduced the Ascend 920 AI chip, which boasts over 900 TFLOPs of processing power and improved energy efficiency. Tom's Hardware+1AIMultiple+1

Emerging Startups

Several startups are innovating in the field of energy-efficient AI:

• EnCharge AI: Developing chips that are up to 20 times more energy-efficient than current market leaders, focusing on in-memory computing to minimize data movement. theaustralian+1Y Combinator+1
• Untether AI: Offers AI inference accelerators with high throughput and superior energy efficiency, suitable for both cloud and edge applications. untether.ai
• EdgeCortix: Specializes in energy-efficient AI processors for edge computing, aiming to deliver high performance with low power consumption. VentureBeat
• Axelera: Focuses on accelerating computer vision on edge devices through a combination of hardware and software platforms. VKTR.com
• Exa Laboratories: Building energy-efficient chips ("XPUs") for AI training and inference, aiming to optimize dataflow and reduce energy usage.

Fifteen leading companies have successfully implemented AI to transform their operations, drive efficiency, and unlock new growth opportunities. By integrating AI across workflows, customer experience, automation, and decision-making, they have demonstrated that AI is no longer a side project but a core enabler of modern business. Their success highlights how thoughtful AI deployment accelerates innovation, enhances productivity, and reshapes industries for a more dynamic future.

1. Airbnb

• Use of AI: Upgraded its Automation Platform to support GenAI with LLM-driven workflows, Chain of Thought reasoning, and advanced context/guardrail systems.
• GenAI Successes:
• Effectiveness: Very mature implementation. Airbnb successfully transitioned from rigid conversational flows to dynamic, intelligent automation.

2. Amazon

• Use of AI: Amazon Q, a GenAI assistant, automates Java version upgrades.
• GenAI Successes:
• Effectiveness: Exceptionally high ROI and operational impact. One of the most successful case studies.

3. CogniSure

• Use of AI: Developed an AI-based system for insurance quote automation using Visual Question Answering (VQA) models and GPT-4o.
• GenAI Successes:
• Effectiveness: Highly successful. Achieved 173% YoY improvement, with GPT-4o performing faster (90s vs. 400s) and more accurately than traditional OCR.

4. Comcast

• Use of AI: Built real-time and post-call churn prediction tools using transcripts, embeddings, LLM summaries, and context clustering.
• GenAI Successes:
• Effectiveness: Smart use of LLMs for real-time customer retention strategies. High impact on customer churn reduction and competitive intelligence.

5. Dick's Sporting Goods

• Use of AI: Built a personalization engine using RAG and LLMs to generate and rank email templates by customer segment.
• GenAI Successes:
• Effectiveness: Strong early results, ready for A/B testing and deployment at scale.

6. Discord

• Use of AI: Built a flexible GenAI product development framework from ideation to MVP and full-scale deployment.
• GenAI Successes:
• Effectiveness: Extremely agile and user-focused. Discord rapidly tests, refines, and ships AI products without over-investing upfront.

7. Grab

• Use of AI: Built an LLM-powered tool called Gemini to automate classification of data, especially PII.
• GenAI Successes:
• Effectiveness: Solved scaling problems and reduced false positives. Currently expanding the solution across business units.

8. L’Oréal

• Use of AI: Developed GenAI as a Service in 3 months to enable secure internal LLM usage.
• GenAI Successes:
• Effectiveness: Rapid deployment with a strong focus on creative use cases and internal security.

9. Pinterest

• Use of AI: Developed a Text-to-SQL feature using LLMs to assist data analysts.
• GenAI Successes:
• Effectiveness: Dramatic performance gains with a 35% improvement in task speed, showing strong ROI on GenAI deployment in analytics tools.

10. Pfizer

• Use of AI: Used RAG and LLMs to classify documents and answer internal knowledge queries.
• GenAI Successes:
• Effectiveness: High-value impact in accelerating drug development and reducing regulatory lag.

11. Takeda

• Use of AI: Streamlined clinical trial protocols using AI to automate SoA (Schedule of Activities) table extraction and harmonization.
• GenAI Successes:
• Effectiveness: Promising early results with high performance and clear applications in protocol optimization.

12. Uber (Case 1: Journey to GenAI)

• Use of AI: Broad application across ETA prediction, Eats, fraud detection, and internal tools using Michelangelo platform.
• GenAI Successes:
• Effectiveness: Significant productivity gains, streamlined AI ops, and scalable LLM integration make Uber a top-tier GenAI adopter.

13. Uber (Case 2: In-House LLM Training)

• Use of AI: Invested in open-source, in-house LLM training infrastructure using Llama2, Mistral, and DeepSpeed.
• GenAI Successes:
• Effectiveness: Technically advanced and highly scalable. Uber is now self-reliant in its LLM deployment lifecycle.

14. Vimeo

• Use of AI: Built an interactive Video Q&A tool using Retrieval-Augmented Generation (RAG).
• GenAI Successes:
• Effectiveness: Highly innovative with strong UX improvements, especially in educational and corporate video use cases.

15. Walmart

• Use of AI: Deployed GenAI across eight functions:
• Effectiveness: Broad application across CX and operations. High maturity and aggressive rollout with clear impact on productivity and seller/buyer UX

AI agents are no longer a future concept, THEY ARE THE NOW. They're actively reshaping business today. Companies like Airbnb, Amazon, Walmart, Pfizer, and others are leading the way, using agentic AI to automate workflows, drive decision-making, and improve customer experiences arent foolish. They're smart and getting smarter. With advances in energy-efficient chips from NVIDIA, IBM, and a wave of startups, the future of AI will be faster, smarter, and more sustainable for our eco-concerns.

Businesses that embrace agentic AI now, autonomous systems that reason, act, and collaborate, will define the next era of innovation. The opportunity is massive, and the transformation has already begun. Stop resisting, stop crying like insolent children, about the current state and help us create the future. Because it literally belongs to those of us willing to create it and tackle the biggest problems and reframe them into opportunities. Change only hurts when you resist it. Embrace it and change the way you look at things and the things you look at will change.

~fin

I'm Thomas. Read my stuff or don't.