Anthropic’s Computer-Controlling Agents Open Door to Real World Uses, Experts Say

Anthropic has unleashed AI agents that can surf the web and control computers, intensifying the race to commercialize AI assistants that can actually get things done.

The move puts Anthropic in direct competition with tech giants Microsoft, Google, and OpenAI, who are all vying to automate routine computer tasks for businesses looking to boost productivity and cut costs. While previous AI tools have focused on text and image generation, these new agents represent an evolution toward AI systems that can actively manipulate software and complete complex workflows with minimal human oversight.

“Agentic AI is set to redefine industries over the next three years, with its ability to operate autonomously and make decisions without human intervention,” Dan Parsons,  COO/CPO at Thoughtful AI, told PYMNTS. “The greatest impact will be seen in areas like system administration, operations, customer service, and complex decision-making, where AI can drastically improve efficiency, reduce costs, and unlock new levels of scalability.”

Anthropic’s new AI capabilities allow its Claude model to control computers directly—from web browsing to coding and app interaction. While the technology shows promise, with Anthropic claiming superior performance on key benchmarks, actual success rates remain modest at 14.9% for basic computer tasks. 

Unlike rivals Microsoft and Salesforce, which build AI agents to work within specific applications, Anthropic’s approach lets its AI system operate computers just as humans do – by interpreting what’s on screen and using the mouse and keyboard. The strategy marks a significant shift in how AI could automate everyday tasks, though experts caution that reliable performance remains a key challenge.

“Right now, the impact of autonomous AI agents is limited, but the potential is very large,” Olga Kupriyanova, principal consultant, AI & data engineering at technology research and advisory firm, ISG told PYMNTS.

The growing competition in the AI agent space will likely drive down AI pricing. According to McKinsey, while AI agents were initially too expensive for many business applications, costs are now rapidly declining. The consulting firm notes that as these technologies become more standardized and industrialized, they’re becoming cost-effective enough to automate even niche business processes that weren’t economically viable before.

“As major companies like Microsoft, Google, and Amazon compete to offer the most advanced AI tools, they’ll need to make their products accessible to a broader market, including smaller businesses,” Arunkumar Thirunagalingam, Manager, Enterprise Data Management at Santander Consumer USA, told PYMNTS. “This competition could result in lower costs, more flexible pricing models (like pay-as-you-go), and bundled services that make AI adoption more affordable. However, smaller businesses may still face challenges in managing the upfront costs of integration and the need for technical expertise.”

Companies eyeing AI to replace human workers face a complex financial equation, Thirunagalingam  said. While AI agents can slash ongoing expenses like salaries and benefits by handling repetitive tasks 24/7, the upfront investment is steep.

Thirunagalingam said that beyond the obvious software licensing fees, businesses must budget for technical infrastructure, staff training, and continuous system maintenance. The true savings only emerge after covering these initial hurdles – making the shift to AI automation a long-term strategic decision rather than a quick cost fix.

“Human workers, on the other hand, provide flexibility, creativity, and emotional intelligence, which AI may struggle to replicate,” he added. “While AI agents excel in efficiency and scalability, human workers are essential for tasks requiring complex judgment, interpersonal skills, and adaptability. Businesses need to weigh these factors based on their specific needs.”

Anthropic’s “computer use” AI agent could transform office work by watching and learning from activity across different software programs. Unlike earlier AI that worked within single applications, this system can track how humans move between various tools to complete tasks, Abhi Maheshwari, CEO at AI agent company Aisera told PYMNTS.

“If an administrative job is mostly about repetitive tasks on the desktop, then this type of role will likely fade away – perhaps quickly. However, this may overstate things,” he added. “Many of these jobs are more than repetitive processes. There is often a need for human interaction – meeting people, having calls, presentations, etc. If anything, there will be more time to devote to these more important tasks.”

The post Anthropic’s Computer-Controlling Agents Open Door to Real World Uses, Experts Say appeared first on PYMNTS.com.