Home » Business Topics » AI Ethics

Use Case Language Models: Taming the LLM Beast – Part 1

  • Bill Schmarzo 

“Sometimes, you don’t know where you’re going until you get there.” – Schmarzo-ism?

Yes, writing this blog turned into a journey. I started in one direction, but after several twists and turns, I ended up with this concept – that use case-centric language models can be combined into entity-centric language models that can support multiple use cases at minimal marginal costs, significantly impacting the economics of AI development.

“I love it when a plan comes together.” – Hannibal Smith, The A-Team

language model is a probabilistic model of a natural language that can generate probabilities of a series of words based on text corpora in one or multiple languages it was trained on

The enthusiasm and intrigue surrounding Large Language Models (LLMs) have exploded (think Oppenheimer-level explosion). In the past nine months, the emergence of cutting-edge Generative AI (GenAI) tools, such as OpenAI’s ChatGPT, Google Bard, and Microsoft Bing, has sparked a surge of enthusiasm and CEO-level interest in LLMs. 

LLMs are robust AI systems trained on massive collections of textual data, enabling LLMs to uncover relationships between subject areas in those massive data sets. The Generative AI’s human-like responses to complex LLM interactions have led to much consternation among industry and government leaders regarding the general dangers of GenAI and general AI.


“Whereas the printing press caused a profusion of modern human thought, the new technology achieves its distillation and elaboration. In the process, it creates a gap between human knowledge and human understanding.”

– Henry Kissinger, Eric Schmidt, and Daniel Huttenlocher, Wall Street Journal

As companies strive to develop and profit from Large Language Models (LLMs), they are encountering challenges associated with the intricate and costly construction and ongoing maintenance of these LLMs. Consequently, companies are exploring Small Language Models (SLMs), which are more focused and contextually specific, and therefore, they are less costly to build, less complex to manage, and provide a clear ROI.

In his blog “Bringing AI to Your Data: The Power of Domain-specific Language Models,” my friend Navin Mukraj discussed the potential of domain-specific small language models:

Domain-specific LLMs are meticulously crafted to cater to sectors like healthcare or finance. These models tap into the goldmine of industry-specific data, presenting insights and solutions that are finely attuned to address the intricacies of niche challenges.

Examples of domain language models include:

  • A travel advice language model that can answer questions and provide recommendations for travelers based on their preferences and needs.
  • A medical diagnosis language model that can analyze symptoms and medical records and suggest possible diagnoses and treatments.
  • A code generation language model that can generate code snippets or programs based on natural language descriptions or specifications.
  • A legal advice language model that can analyze contracts and extract the key terms, clauses, and obligations for the parties involved.
  • An educational support language model that can generate questions based on a given text or topic and grade the answers provided by the students
  • A financial advice language model could help with budgeting, investing, saving, etc. It could use data sources such as income, expenses, assets, liabilities, goals, etc. It could generate outputs such as financial plans, recommendations, reports, etc.
  • A diet advice language model could help with nutrition, fitness, weight loss, etc. It could use data sources such as calories, macronutrients, food preferences, allergies, etc. It could generate outputs such as meal plans, recipes, tips, etc.

Domain language models are a more cost-effective option for companies seeking to exploit the language model explosion. However, I believe organizations can take one more step towards leveraging and swiftly capitalizing on the language model trend – use case-specific small language models (use case language models).

Use Case Language Models

Use case language models are trained or adapted to excel in solving specific, cross-domain problems by integrating and analyzing pertinent data from various data sources to deliver comprehensive and actionable responses, such as improving customer retention, increasing sales revenue, or optimizing inventory management.

Use case language models have several advantages over LLMs and domain-centric small language models, including:

  • Accuracy: provide more accurate and relevant answers for specific problems or goals across multiple domains. They can integrate and analyze data from multiple data sources that are relevant and informative for that problem or goal.
  • Efficiency: process natural language inputs and outputs faster and more efficiently, as they require less computation and memory resources. They can also run on devices with limited resources (smartphones, IoT devices) without compromising performance or functionality.
  • Interpretability: explain their natural language outputs and provide evidence or justification for their decisions or actions. They can also allow users to understand their internal logic and reasoning, which are more transparent and understandable.
  • Controllability: allows users to adjust their natural language outputs and modify their behavior according to their preferences. They can also better adhere to ethical and social norms, such as privacy, fairness, safety, etc., as they are more accountable and responsible.

Examples of use case language models include:

  • A customer retention language model to retain customers and reduce churn. It would integrate and analyze data from customer feedback, sales records, market trends, etc., and generate personalized offers, recommendations, and customer incentives. For example, a customer retention language model could help a subscription-based service identify customers at risk of canceling their subscriptions and send them personalized messages or emails highlighting the benefits of staying with the service based on their usage patterns or offering discounts or free trials.
  • An inventory optimization language model to optimize inventory levels and reduce costs. It would integrate and analyze demand forecasts, supply chain information, product specifications, etc., and generate optimal business inventory plans, orders, and allocations. For example, an inventory optimization language model could determine how much inventory to order for each product category and location based on historical sales data, seasonal trends, customer preferences, etc.
  • A marketing campaign effectiveness language model to improve the effectiveness of marketing campaigns. It would integrate and analyze campaign performance metrics, customer behavior data, market research data, etc., and generate actionable insights and recommendations for the business. For example, a marketing campaign effectiveness language model could evaluate marketing campaigns’ return on investment (ROI) across different channels and platforms, such as email, social media, web, etc.

How can we ensure the relevant use of use case language models? We can embed them as a component of a larger solution that addresses a specific business or operational challenge that spans multiple domains.  We can embed them into our AI-powered Apps.

Integrating Use Case Language Models into AI App Development Canvas

AI Apps are domain-infused, AI/ML-powered applications that continuously learn and adapt with minimal human intervention in helping non-technical users manage data and analytics-intensive operations to deliver well-defined operational outcomes.

By integrating the Use Case small language model into our AI Apps, we enable our stakeholders to utilize the language model’s data analytics capabilities for improved decision-making or problem-solving. The Use Case language model further enhances the capabilities of our AI app by enabling stakeholders to explore and analyze data more efficiently.


Figure 1: Updated AI App Development Canvas

Let’s say that we want to build a “Customer Retention” AI App for marketing, sales, and customer service to identify, segment, and target customers at risk of leaving or with high growth potential and create personalized offers to deliver the following results:

  • Improved customer retention rate by predicting and preventing customer churn with timely and effective interventions.
  • Increased customer loyalty rate by rewarding and recognizing loyal customers with incentives and referrals.
  • Enhanced customer growth rate by upselling and cross-selling relevant products and services to high-potential customers.
  • Reduced operational costs by automating and optimizing customer retention and growth campaigns.

We could integrate a “Customer Retention” (use case) language model into our Customer Retention AI App that would support the business stakeholder’s engagement and interrogation of the Customer Retention AI app.  That Customer Retention language model would be built from the following data sources: Sales transactions, Customer support, Customer demographics, Product returns, Payments, and Social Media.

Use Case Language Models Summary

Use case language models are more focused, more cost-effective, and can provide a more immediate ROI than LLMs.  They become a key enabler in creating AI-powered apps that can drive more meaningful, relevant, responsible, and ethical outcomes.

However, I can enhance the economics of language models by transforming Use Case language models into Business Entity language models.  Let me explain in Part 2…