New AI era defined by agents, rising costs and maturity gaps

The rapid evolution of AI models, growing enterprise interest in agentic systems and mounting concerns over cost management are reshaping the technology landscape, according to Gartner AI experts Arun Chandrasekaran and Pieter den Hamer, who warn that organisations remain far less mature in AI adoption than industry hype suggests.

AI innovation is accelerating while enterprises continue to grapple with governance, adoption and return on investment challenges.

The analysts said competition among leading AI vendors is driving unprecedented innovation, but the advantage gained from releasing the most capable model is becoming increasingly short-lived.

"The leaderboards keep changing every quarter," Chandrasekaran said.

As a result, vendors are converging around several key areas of competition: model efficiency, reasoning capabilities, multimodal functionality and the emergence of AI agents capable of orchestrating workflows across enterprise systems.

One of the most significant developments is the rise of computer-use capabilities and tool integration, which are enabling AI systems to perform increasingly complex tasks on behalf of users.

Advances in speech recognition and reasoning models are creating new opportunities in customer service environments - specifically in the deployment of voice agents.

The growing sophistication of these systems means that customers contacting airlines, banks, insurers or other service providers may increasingly interact with AI-powered agents before speaking to a human operator.

East-west divide emerging in AI models

There is a growing distinction between AI innovation in Western and Asian markets.

While leading US vendors continue to focus largely on proprietary models, many of the most capable open-weight models are emerging from China, including offerings from companies such as Alibaba, DeepSeek, MiniMax, Moonshot and Zhipu AI.

Chandrasekaran has observed widespread experimentation with these models across APAC, in both pilot and production environments.

The trend could have significant implications for enterprise AI adoption as organisations seek alternatives to proprietary frontier models and look to reduce deployment costs.

SaaS disruption is real - but no apocalypse 

Speculation that AI agents could fundamentally disrupt the software-as-a-service industry continues to grow.

The Gartner analysts discussed how AI will reshape SaaS economics, but were quite adamant that the touted SaaSpocalypse was unlikely to actually come to fruition.

Large SaaS providers continue to possess significant advantages through deep workflow integration, extensive domain-specific data, regulatory compliance capabilities and established customer relationships.

However, their traditional seat-based pricing models are increasingly being challenged by agentic AI.

As organisations deploy AI agents capable of performing work independently of human users, customers are beginning to question why software pricing remains tied to employee headcount.

This is driving interest in alternative commercial models, including consumption-based, workflow-based and outcome-based pricing.

The shift creates a difficult balancing act for established software vendors, which must satisfy customer demands for more flexible pricing while protecting margins and recurring revenue streams.

Customer service is emerging as one of the earliest examples of this transition, with some vendors experimenting with pricing linked directly to customer satisfaction scores and business outcomes rather than user licences.

AI costs becoming a boardroom concern

While much of the industry conversation remains focused on AI capabilities, economics are becoming a critical issue for enterprise buyers.

Rising token consumption, increasing use of AI agents and evolving pricing models are creating significant uncertainty around long-term costs.

Unlike traditional chatbots, autonomous AI agents can generate large volumes of requests to underlying models, dramatically increasing infrastructure utilisation.

"The AI free lunch is starting to not look so free anymore," Chandrasekaran said.

The transition from subscription-based pricing toward consumption-based charging is adding further complexity.

Many organisations lack the monitoring tools, governance frameworks and financial controls needed to accurately forecast AI spending.

But token costs represent only part of the challenge. Data integration, governance layers, model monitoring and employee training all contribute to the total cost of ownership.

As enterprises move beyond pilot programs, many are discovering that AI budgets are more difficult to predict than traditional software investments.

Most organisations still in middle of maturity curve

Despite widespread AI adoption, Gartner's research suggests relatively few organisations have successfully scaled AI across their operations.

Only 17 per cent of organisations have reached what the firm classifies as a high level of AI maturity, Den Hamer revealed.

These organisations have embedded AI broadly across business functions and achieved enterprise-scale adoption.

A further 51 per cent fall into a medium-maturity category, where AI is being used in isolated projects and departments but has yet to deliver transformational impact.

Around one-third (32 per cent) remain at low maturity levels, focusing primarily on experimentation and proof-of-concept initiatives.

The findings suggest a substantial gap between AI enthusiasm and operational reality.

The same pattern is evident in agentic AI adoption.

Although agents have become one of the most heavily marketed technology categories of the past year, Gartner estimates only around one in five organisations currently have agentic systems running in production environments.

"The remaining 80 per cent is only experimenting with agentic AI or is just thinking about it," Den Hamer said.

AI literacy emerging as a competitive advantage

The organisations achieving the strongest AI outcomes share several common characteristics.

Mature adopters are using AI to accelerate research and development, improve operational resilience and enhance product quality, as well as redesigning business processes around AI, not simply layering new tools onto existing workflows.

Most importantly, they are investing heavily in workforce education.

Den Hamer continues to observe a strong correlation between AI literacy and business value.

"Everyone needs to learn about AI," he said.

"If you do that proactively, we clearly see that ROI then tends to be much higher compared to not (educating employees)."

Organisations must also address employee concerns about job security if they hope to accelerate adoption.

A culture of psychological safety, combined with clear communication about AI's role within the workforce, will be essential for scaling deployments successfully and building productive workplace culture around automation and emerging tools.