Your Company’s Next Killer App is a “CLM” (Custom Language Model)

Doing business at the speed of AI is no easy feat—the developments related to integrating artificial intelligence-powered technology change daily and weekly, and sometimes, it feels almost hourly. Still, the infrastructure and architecture for AI-powered experiences are indeed being built. Apple most recently signaled that it is in talks with Google to integrate Google's LLM, Gemini, into the Apple OS. Rumors are circulating that Open AI could also be a contender.

Either way, what Apple is doing signifies the consumer version of the same trend we’ll be seeing on the Enterprise side of things—a company’s conversational experiences are going to become more intelligent, more data-driven, and feel a lot more human. Industry luminaries following AI will call out the potential of “AI Agents”—a not-too-distant future scenario where AI won’t just be your co-pilot but will integrate into technology ecosystems to help initiate, perform, and complete complex tasks—like booking a family vacation or managing your medical care. But, while the vision for AI Agents is both right and ambitious, we’re skipping over some key transformations that need to get started before beginning to lay the groundwork for a future where AI Agents will be part of the customer/employee/patient experience.

CLMs Will Act As A Bridge Between Yesterday’s Chatbots and Tomorrow’s Digital Experience?

CLMs or Custom Language Models will help bridge the gap between today’s status quo, where a company either has “dumb” chatbots in place that are limited in functionality or doesn’t have them at all. The distinction between a traditional Chatbot and one that leverages a language model is becoming more clear as users become more familiar with the kinds of experiences they can now have on platforms like ChatGPT, Perplexity, Pi, Gemini etc. It’s why Apple is choosing to partner to replace Siri and is entertaining partnerships—vs. going at it alone. Users are quickly becoming accustomed to the breadth of conversational content they can engage on these platforms, and the expectation will soon be: “Shouldn’t everything work this way”? However, many companies are currently working through several challenges as I write this in the pursuit of creating intelligent conversational experiences that can help set themselves up for a world where AI agents will become a seamless part of the customer/employee experience. They are:

Data security: Building a CLM requires sharing proprietary and potentially sensitive data with AI-fueled large language models—the key concern here is training somebody else’s AI on your proprietary data. As outlined in the above visual—enterprise-level security in servers and how third-party technology (Google PaLM, for example) can be set up to meet enterprise data security standards.

Guardrails: Large Language Model (LLM) engines require prompting, fine-tuning, and maintenance to ensure they don’t “hallucinate,” potentially putting your corporate reputation in danger or exposing the company to liability.

Maintenance: Any deployment fueled by a language model is a significant commitment and will require regular maintenance, both in updating data and ensuring the quality of output. While AI has the potential to automate processes, services, and systems, it is not without significant human assistance, including maintenance.

CLMs Will Redefine Your Customer’s Relationship WIth Data

As Google already knows all too well, our relationship with data is rapidly being redefined. For example, the traditional data search experience needs to be updated in real time. Entering search keywords and then manually scouring Web pages, documents, or individual videos will be different from how we interact with large amounts of data even in the near future, and Google knows this. However, the same goes for the millions of digital experiences companies have constructed for their end users.?

The data science team I work with has been evolving this kind of traditional search experience with a company with millions of documents in its database as its business model. By constructing a CLM, the legacy experience of searching for documents, downloading them, or having to watch full videos to find the exact content you’re looking for will be replaced by a multimodal conversational interface:

Removing the friction of downloads and manual sifting through content to find exactly what you’re looking for conversationally and contextually is already changing how we view data and information. These experiences are already happening as consumers utilize AI apps and services, and as we’ve learned from experience, the consumerization of IT is inevitable.?

CLMs are coming to a boardroom near you. Your CEO is probably already asking about it—you may already be experimenting. You might be calling it something else (in the past, I’ve called this “Enterprise LLM”). But whatever you call it—Apple, Google, Microsoft, and other tech players are actively re-building the digital experiences we all use to become more contextual, conversational, and intelligent. We’re looking at a near future where those technologies have set a new standard, and our digital experiences provided by the company/enterprise feel outdated. CLMs will act as that first step forward.