We’ve all used Large Language Models (LLMs) like Perplexity, Llama, Gemini, ChatGPT, Claude, and others, which are designed to comprehend, interpret, and generate human-like language. My current favorite is Perplexity.ai…but that could change as other models fight for ‘AI dominance’. 

Prediction: It’s my humble opinion that these models will eventually become commoditized (i.e., anyone will be able to have them and/or own them). 

If so, then... 

For companies who want to find a competitive edge in an AI world, developing Specialized Large Language Models (SLMs, pronounced ‘slims’) focused on a specific market domain is the MOVE to make. 

The clearest indication that this is already happening and at scale is how aggressively the Business Process Outsourcing (BPO) market is moving in that direction. Outsourcing is a huge business globally (~300 billion and growing) as BPOs handle tasks like customer service or support (i.e., call centers), payroll, invoicing,…etc. 

It should be noted that enterprises usually choose to outsource because of the cost of doing it in-house, or they simply don’t have the expertise to do it.  

Keep this in mind as it will become relevant towards the end of this article!

With the rise of rise of AI Agents, (read my book; shameless plug) BPOs are now leveraging domain-specific AI agents to transform processes across different industries. At the beginning of this year, INFOSYS announced that they are developing over 100 new GenAI Agents for its clients. 

Agents in:

- Healthcare, vertical AI agents can automate claims processing by analyzing patient records and insurance protocols.

- Banking, these agents streamline compliance checks and fraud detection.

- Retail, generative AI-powered agents personalize recommendations.

These specialized AI Conversation Agents are now being deployed. Infosys’ focus on specialized vertical AI agents underscores the AI transformation in the BPO market. 

But, there is a danger here! 

As AI becomes more commoditized and the models easier to train without having AI experts, what’s to stop companies from bringing sLLMs in-house if they can mitigate the cost and the ease of development makes it possible for them to do it themselves? 

Hmmm…

If I ran a BPO company like InfoSys,…this would keep me up at night!