By Sumedh Barde, CPO of Simbian

LLMs are revolutionizing the entire technology industry. They have improved a lot while lowering the cost. In 2025 we will see leading LLMs start to offer their services for free. Companies like OpenAI and Anthropic may decide to give away their leading models for free while monetizing only the advanced features. Some of our predictions for 2025:   

LLMs hallucinate and are unreliable. They will confidentially and silently give a wrong answer. LLMs were supposed to automate security, but this inaccuracy posed an existential threat. Even a simple mistake can create havoc for an organization. If AI shuts down the wrong application, or blocks the wrong network port, or gives access to the wrong employees, the result can be extremely damaging. 

The next generation of AI tools beyond LLMs are "AI Agents," which take actions autonomously. Domain experts, such as in security operations, can build domain-specific agents. Such agents abstract LLMs from the end users and are autonomous. 

In 2025 we will begin to see mainstream adoption of AI Agents across enterprises. Enterprises will find AI Agents are a pragmatic way to augment teams such as security operations, human resources, governance, and internal helpdesk. These teams are chronically understaffed, so AI Agents provide relief.  

The quality of AI Agents depends on the data they were trained with. With the rise in personalized AI models and Agents, it will no longer be possible to gate the training process with human data scientists. Data security companies focus on access to data historically; but quality and correctness of the data will become more important than ever. 

The internet commoditized the middleman in commerce by enabling suppliers to connect with buyers anywhere, whether it's in goods or finance or education. GenAI takes that to another level by unlocking opportunities for the global south, as well as for underprivileged communities in the west. Until now, to succeed as a supplier, it was not enough to produce the goods. You needed a team good at marketing, support, in some cases writing software etc. Or you could outsource these to experts, if you had the money.

GenAI breaks this barrier to entry. It is significantly easier to create software, create marketing artifacts, communicate across cultures etc. Anyone who has the domain expertise and the ability to articulate clearly, now has a virtual team at their disposal.  

Today, Google is free but Google Cloud Service is a paid service. LLMs may follow a similar path.  2025 will bring fundamental shifts in how AI will be used across the globe, LLMs will be commoditized.

##

ABOUT THE AUTHOR

Sumedh brings over 15 years of product and engineering experience in security and digital rights management. Prior to Simbian, Sumedh was Head of Product at Microsoft for Azure’s data security products, and Director of Security programs at Meta. In these roles, Sumedh led the teams building some of the world’s most security-critical and high-scale cloud services, now used broadly across enterprises.