Anthropic Says Companies Should Use AI for More Than Efficiency

 A good next story is Anthropic's warning that companies should use AI for difficult problems, not just efficiency. It has a strong business + AI angle and is different from the UPI stories.

Companies should stop looking at artificial intelligence only as a way to make existing work faster and cheaper, according to Anthropic India Managing Director Irina Ghose.

Speaking at the Global Fintech Fest 2026, Ghose said businesses should increasingly use advanced AI models to tackle problems that have traditionally been difficult or time-consuming for humans to solve. She pointed to areas such as drug discovery and understanding old banking software as examples. 

AI is moving beyond simple efficiency gains

A large part of corporate AI adoption so far has focused on productivity. Companies are using AI to summarise documents, write code, analyse information and automate repetitive tasks.

Ghose argued that this is only the beginning. According to her, the bigger opportunity comes when businesses give frontier AI models problems where the answer is not already obvious and where traditional approaches can take months or years.

That could include scientific research, complex engineering work and highly specialised business problems.

Drug discovery could be one major use case

Healthcare is one area where advanced AI could have a particularly significant impact.

Drug discovery can require large amounts of research, experimentation and analysis before a promising candidate moves forward. More capable AI systems could help researchers process information and explore possibilities that would otherwise take much longer.

Ghose used drug discovery as an example of a problem where the capabilities of frontier AI could be more valuable than simply automating routine office work. 

Banks could use AI to understand old software

Another example highlighted by Anthropic is the banking sector's legacy technology.

Large financial institutions often rely on software systems that have been running for many years. Understanding, updating or replacing these systems can be difficult because the original code may be complicated and documentation may be incomplete.

Advanced AI models could potentially help developers understand these older systems and speed up work that traditionally requires specialised expertise.

AI models are becoming cheaper to use

The push towards more advanced applications is also being helped by falling AI costs.

Ghose said Anthropic's latest models are around 25% cheaper for typical workloads, while more complex, agentic tasks can see savings of up to 45%. Lower costs could make it more practical for companies to use powerful AI models for tasks that previously would have been too expensive.

As AI becomes cheaper, businesses may have more room to experiment with applications beyond chatbots and basic automation.

Companies need to rethink how they measure AI

Another challenge is that businesses often evaluate AI using traditional measures such as hours saved or reduced operating costs.

Those measurements are useful, but they may not capture the value of solving a problem that was previously considered too difficult or expensive to tackle.

For companies, the bigger question could therefore become less about “How much time can AI save?” and more about “What can we now do that wasn't practical before?”

AI alone will not guarantee a competitive advantage

Ghose also cautioned that simply adopting frontier AI will not automatically give a company an advantage.

As increasingly capable models become more widely available, businesses will need other strengths to stand out. Distribution, customer trust and proprietary data could become important differentiators.

That means the companies that benefit most from AI may not necessarily be those with the biggest models, but those that know how to apply the technology to meaningful problems.