A Google GenAI expert weighs in on why companies are clamouring for AI agents

Google, doubling down on artificial intelligence (AI) to fuel revenue growth, is witnessing increased enterprise interest in agentic workflows, according to Oliver Parker, vice president of global Generative AI (GenAI) go-to-market at Google Cloud.

In the last few months, the interest has been growing significantly, Parker noted during a recent interview in India.

AI agents—autonomous systems capable of decision-making and action—are designed to perform tasks with minimal human intervention. From driverless cars to smart home assistants and trading bots, these agents represent a critical evolution in AI technology.

Read this | AI and GenAI are transforming robotics: ABB robotics global head

Google’s newly launched AgentSpace platform enables businesses to create AI-powered workflows that integrate seamlessly into operations, offering tools for tasks like enterprise search and data interaction. It is “taking AI beyond developers to the business layer,” Parker explained.

Over the past year, Google has poured resources into developing agentic models. These systems can process broader contextual data, anticipate multiple steps ahead, and execute tasks autonomously under human supervision. CEO Sundar Pichai underscored this shift in a December blog post announcing Gemini 2.0, describing it as the foundation for the “next era of models built for this agentic era.”

Building agentic systems in low-code environments is emerging as the quickest way to drive enterprise value.

When to use AI agents?

The value of agentic systems lies in their ability to trade off latency and cost for enhanced task performance, and those using these systems should consider when this tradeoff makes sense, noted a 20 December blog by tech company Anthropic.

However, companies need to deploy these systems judiciously, Parker said. “Think of a model as an engine and an agent as a car,” he explained, emphasizing that many enterprises embed large language models (LLMs) into workflows without creating fully autonomous systems.

For cost-sensitive and latency-critical use cases, Google offers Gemini Flash, a model optimized for speed and affordability. In contrast, Gemini Pro caters to reasoning-intensive tasks, albeit at higher costs and slower response times.

Parker cited a Southeast Asian chat assistant as an example of Gemini Flash’s utility, while Gemini Pro excels in scenarios requiring complex decision-making.

Google’s AI strategy extends to Vertex AI, its enterprise developer platform that supports diverse models. Partnerships with Anthropic, Cohere, Mistral, and Meta’s Llama further underline Google’s commitment to flexibility, offering enterprises a suite of AI tools tailored to their needs.

Enterprise adoption of AI is accelerating globally, despite concerns about return on investment (ROI). Parker observed that while US enterprises focus on extracting value from business platforms, countries like India and China—traditionally developer-centric—are increasingly embracing higher-level services. “Building agentic systems in low-code environments is emerging as the quickest way to drive enterprise value,” he said.

Success stories such as Apollo 24/7 Health’s AI-powered consultation assistant highlight the transformative potential of these solutions. These systems streamline processes and enhance user experiences while driving down costs.

Read this | How AI, GenAI malware is redefining cyber threats and strengthening the hands of criminals

“Over the next 12–24 months, we’ll see progress in models and platforms, with enterprise value shifting beyond developers to packaged solutions like agent-based platforms,” Parker said, citing AgentSpace’s seamless integration with tools like ServiceNow, Workday, and Salesforce.

Tackling rising competition

However, competition in this space is heating up in the cloud infrastructure services space, which includes infrastructure-as-a-service (IaaS), platform-as-a-service (PaaS) and hosted private cloud services. Third-quarter enterprise spending on cloud infrastructure services was $84 billion worldwide, up $15.7 billion or 23% from the third quarter of 2023, according to the Synergy Research Group, with generative AI being a major factor behind the market acceleration.

That said, Amazon maintains a strong lead with a 31% market share, followed by Microsoft (20%) and Google (13%). Among the tier two cloud providers, those with the highest year-on-year growth rates include Oracle, Huawei, Snowflake and Cloudflare, according to Synergy Research.

Parker countered, “Google’s vertically integrated stack—from silicon to infrastructure—provides a unique advantage, allowing us to optimize for cost, latency, accuracy, and performance. This approach enables tailored solutions across diverse use cases, such as leveraging long context windows (up to 2M tokens in Flash)”.

Google, according to Parker, anticipates a shift toward AI-first tech stacks, “which may redefine the hyperscaler (big cloud services providers) landscape as companies prioritize AI partnerships over traditional infrastructure”.

He added that Google has made “significant improvements over the past year, reducing errors and enhancing reliability”. For regulated environments, human oversight remains crucial. Tools like grounding with Google Search enhance accuracy, particularly for currency and factual checks, “although no model is flawless yet”, Parker asserted.

As an employee myself, I use tools like our Gemini system daily, showcasing the importance of adopting advanced technologies.

Google, he added, is also championing energy efficiency–from water cooling data centres to optimizing tensor processing units (TPUs), “maintaining leadership in sustainability while meeting AI compute demands.”

Ongoing learning

But what skills will employees need in this era of high automation?

“As an employee myself, I use tools like our Gemini system daily, showcasing the importance of adopting advanced technologies,” Parker said. He acknowledged that skilling the workforce is critical—not just for technical teams but across the board.

Parker also underscored the importance of multimodal AI, like the company’s Gemini models, which integrate text, speech, images, and video. “Over 80% of recent meetings highlighted the growing demand for such capabilities, which open up new use cases and solidify AI’s role in enhancing human productivity,” he said.

Training, according to Parker, isn’t just about building AI but also using it effectively. “For instance, in 2023, we committed to training 1 million people in two years, but we achieved that goal within 12 months by leveraging platforms like YouTube and targeting diverse audiences—startups, traditional workers, and beyond,” he added.

Also read | Five artificial intelligence trends to track in 2025

Acknowledging that the pace of change is challenging, Parker said Google is adapting by rapidly developing content and using its vast resources to distribute it. Education helps people see AI as an economic benefit rather than a threat to their skills, he concluded.