Right Prompts for Using AI for Your Business – Introducing MathCo’s Ask NucliOS

Article
By
Adithi Pai
Krishna J Nair
December 2, 2024 9 minute read

In an age where technological advancements and transformations occur at unprecedented rates, data and AI have become indispensable assets for businesses. Organizations churn out and generate a sea of data on a regular basis, but there aren’t enough hands on deck to swift through them and identify quality insights. Subsequently, the true potential of data is often masked or unrealized, leading to considerable loss in value generation.   

Addressing this pitfall requires nuanced skills, knowledge of multiple data sources, and effective collaboration between multiple stakeholders. The application of AI brings a combination of these factors, and the recent advancements in Large Language Models (LLMs) open multiple avenues for insights generation by asking the right set of questions. An adequately trained AI-powered assistant can derive insights from untapped large data sets within seconds, aiding teams and enhancing efficiency. They can also empower businesses to become future-ready by analyzing data and predicting future trends, enabling personalization, and automating manual processes, thereby freeing bandwidth for business functions.

Ask Away with Ask NucliOS   

The industry has been witnessing a sudden boom in smart assistants with various offerings and abilities that promise to transform processes. There has been significant widespread adoption of copilots such as Microsoft Copilot, Einstein Copilot, Github Copilot, and more. While they continue to spark innovation and are excelling at solving specific business problems, enterprises need custom copilots connected to multiple sources of information, or literally a pilot that understands business context and goals and can take them to the right destination seamlessly. 

At MathCo, our ultimate aim is to empower our clients to own their intelligence and take complete control of their data. With Ask NucliOS, our proprietary conversational assistant platform, we equip our clients to develop their ultimate AI assistant. Built for flexibility and customization, users can configure their AI assistant and connect to relevant data sources, adding adequate organizational context to generate reliable answers accessible through a customizable interface. Some of the AI assistants we have developed with our clients include:  

  1. SalesRep assistants that help pharmaceutical representatives prepare for potential meetings  
  2. Data discovery assistants that aid in finding the right data from a range of SQL tables  
  3. Insights assistants that are integrated with existing dashboards, tuned to answer complex queries   
  4. Brand insight assistants that equip businesses to speed up market research for healthcare companies   

Ask NucliOS opens a world of possibilities for enterprises to develop a model that is one of a kind, built solely to solve their business challenges.   

A Unique Framework Built for Customization 

What makes Ask NucliOS stand apart from the rest of the platforms available in the industry are its core concepts, which include:  

  1. Onboarding custom skill sets that are fine-tuned to perform specific tasks  
  2. Orchestrating the right skill sets needed to answer a user’s question  
  3. Leveraging custom context to generate relevant answers   
  4. An intuitive interface that provides trusted responses, which can also be personalized.   

Custom Skill Sets   

MathCo’s experience with numerous enterprise assistant deployments helped us understand what the industry has been lacking and aided us in creating a Reusable Skill Set Marketplace. This helps us simplify the process of creating assistants as the user does not have to start from a clean slate but rather leverage a pre-existing base to build their solution. Our repository contains advanced skill sets that are readily configurable. With this feature, enterprises can select their preferred language model and connect their skill sets to relevant data sources. Some of the most in-demand skills include:   

  1. Text to SQL Generator: This skill set bridges the gap between natural language and database querying, enabling users to effortlessly translate plain English into SQL by using the schema information and providing integrated visualizations for actionable insights. 
  1. Document Querying Pipeline: Our enterprise RAG pipeline is equipped with configurable LLMs and data connectors to file-sharing sources like SharePoint. This pipeline includes preprocessing layers for PowerPoint documents, enabling the extraction of charts and smart art to prepare them for LLM processing.   
  1. PowerPoint Storyboarding: This skill set is used to automatically create PowerPoint slides from conversations allowing our users to export results outside of our platform for their consumption.  

Our continued collaboration with key industry players helps us expand this marketplace with new offerings and onboarding alternate versions of existing skill sets for specific use cases. Additionally, our platform allows users to publish their unique skills to the Skill Set Marketplace efficiently for easier access and integration for their projects. 

The Brain   

MathCo’s approach to building an AI-powered business assistant includes developing an LLM-powered brain that has distinct abilities that most AI assistants lack or fail at. These include:   

  • Understanding the right question and seeking user input or intervention wherever necessary.   
  • Utilizing only client data and adding organizational context to the query before generating a response to improve accuracy and reliability, thereby reducing the possibility of hallucinations.  
  • Leveraging the right data sources that contain the answers to the queries.   
  • Generating responses that are consumption-friendly for human users. 

While having a skill set is the first step, it is in how the skill set is leveraged and deployed where the success of that skill set lies. Ask NucliOS has taken two steps further and beyond to execute this by including the ability to route queries to the appropriate skill set based on the user query.  

For instance, if a user asks a question that requires pulling data from a SharePoint knowledge base, the orchestrator routes the query to the skill set specialized in handling document queries. If another query involves executing an SQL command to retrieve data and visualize it, it directs the question to the Text to SQL skill set.  

To ensure that queries are routed to the right skill set for the task based on user queries, our platform offers two orchestrator options built for scalability:  

  • Description-Based Orchestrator: This orchestrator uses a prompt template to interpret user queries and route them to the appropriate skill set. It’s ideal for scenarios where the assistant needs to handle varied tasks and requires accurate routing based on user input.  
  • Function Call Orchestrator: For more complex queries, this orchestrator engages users to provide specific parameters needed to execute certain tools, ensuring that even ambiguous or multi-part questions are handled effectively.  

Both orchestrators are designed to support multiple models, enhancing flexibility, optimizing costs, and allowing organizations to tailor the chatbot’s behavior based on the task at hand. Our platform offers a unique mix-and-match model selection feature, allowing users to choose from in-house, open-source, or customized models using the LLM workbench. Most popular models are already onboarded and ready for use.   

A foundational brain that processes key information can help businesses solve almost all complex problems seamlessly and even equip the business to be future-proof. Consider the below illustrative example:  

Our client, a key player in the CPG industry, has an additional budget of $1 million that can be spent in the next quarter. They want to understand which promotions can be run within this cost to achieve an incremental sale of $3 million. 

Ask NucliOS answers this optimization question by considering the client’s internal business workflow, acquiring knowledge of the client data and KPIs, and leveraging an optimizer to generate relevant results through the following processes.   

 

Customizable Contextual Layer  

Though we developed Ask NucliOS after thorough research and industry analysis, our successful deployments result from integrating organizational context to LLMs to generate relevant results. This ability is what sets NucliOS apart from other platforms. With organizational context and data accessible, LLMs can generate reliable answers that can help businesses accelerate their processes and decision-making cycle.  

This manifests as deep understanding of context like metadata information, user persona and roles, metric definitions and even complex business workflows; without which LLM generated answers are not always insightful.   

Ask NucliOS helps users add customer context to their assistants through a dedicated configurator. It contains multiple templates to onboard context. This allows the skill set and orchestrators to fully leverage organizational context, providing more accurate and relevant responses. Additionally, the platform is designed to maintain and build upon the context throughout a session or over multiple interactions, enabling it to handle complex, multi-turn conversations with nuanced understanding.   

An Entrusted AI Assistant to Lift Your Business 

In an era where AI is often accused of hallucinating and misinterpreting data, it is often a challenge to trust AI generated insights and move forward. Ask NucliOS bridges the trust barrier by making the process for generating the final answer transparent, hence ensuring the responses to be reliable. Some of these processes of Ask NucliOS are enlisted below.   

  1. Users can generate citations with our platform to check whether the information used comes from documented sources or is generated from data, such as SQL tables. By seeing what goes under the table, the user can know exactly what empowered the generated responses.   
  2. If additional context is fetched from the contextual layer to generate a certain response, the context is revealed along with the answer.   
  3. Our skill sets include checks and configurable layers to validate responses and discard irrelevant queries, enhancing response accuracy by integrating additional context into LLMs.  
  4. We offer automatic evaluation of RAG (Retrieval-Augmented Generation) pipelines, allowing LLMs to generate test data and provide industry-standard metrics.   
  5. We continuously collect user feedback, which Ask NucliOS leverages to refine and improve the responses. This helps the platform adapt to user needs.   

Ask NucliOS is a platform that is truly yours to believe in and work with, and this is further evidenced by its highly customizable interface. Ask NucliOS can adapt to your brand with a few clicks that enables it to align with your brand guidelines, goals, and more. Your one-stop AI solution can no longer be simplified than this, and it starts with the MathCo Way!  

Keen to know how you can build your own advanced AI assistant with Ask NucliOS and build your ultimate agent who can tackle your business challenges? Get in touch with us for a demo here.   

All

Redefining Enterprise Transformation: The Power of Generative AI

Read more
CPG

3 Ways to Unlock CPG Supply Chain Monetization through AI-led Digitization

Read more
All

Breaking Boundaries with Generative AI: A Technological Revolution

Read more