Nov 14, 2023

An inside look into Gaspar AI’s Generative AI helpdesk

An inside look into Gaspar AI’s Generative AI helpdesk

Share this article

Generative AI for the enterprise is becoming a staple. When looking for potential solutions to automate your employee helpdesk, it’s important to understand their AI technology and features as the issue resolution rate and speed, as well as the employee experience and adoption, depend on them.

Gaspar AI’s generative AI helpdesk is not just another platform—it's a game-changer. It’s been developed with the end user in mind: leveraging advanced AI models, we built a helpdesk capable of not only holding fluent dialogues but also solving complex enterprise challenges end-to-end. Our Generative AI helpdesk was created with three principles in mind:

  • Compelling interactions: We aim to enhance every interaction, making them more engaging and intuitive than ever. With our advanced conversational AI, we ensure that employees enjoy using our platform every day. This engaging experience results in more inquiries, task completions, and problem resolutions.

  • Accelerated results: When employees encounter obstacles, they rely on our helpdesk to swiftly guide them back to productivity. We strive for seamless interactions, ensuring minimal friction: by streamlining dialogues, minimizing steps, and harnessing powerful AI capabilities, we expedite task fulfillment.

  • Infinite possibilities: We aspire to eliminate any artificial limitations on what our helpdesk can do, so that it supports as many use cases as possible across various systems and teams. Regardless of the workflow complexities or business systems, our helpdesk stands ready to meet the challenge. Powered by generative AI, it is fully scalable and continuously expands its knowledge and skills.

In our previous blog post we discussed our platform’s technology behind automatically solving employee requests. Now, guided by the three principles mentioned above, I'd like to highlight the essential processes that enable our helpdesk to provide this smooth, effortless experience, and the building blocks that power up our Artificial Intelligence engine.

Gaspar AI’s Artificial Intelligence methods and processes for best-in-class support

Our helpdesk follows several AI methods and a very thorough process to serve our customers with the best solutions:

Classification

Gaspar AI's intelligent engine classifies user queries – that's the first step towards optimum issue resolution. Once our helpdesk receives an employee request, it identifies its type: whether it requires automation, a knowledge resource-based answer, or a personalized info response. Our classification method is the backbone of seamless user experiences, ensuring that all employee queries are correctly classified so that the right solution is provided.

Information Retrieval

We've harnessed the power of state-of-the-art Information Retrieval methods to fetch the most relevant information and documents from our customers’ knowledge base or internal resources. Our platform integrates with various knowledge bases (such as Coda, Notion, Google Docs, Confluence and more) to ingest and retrieve the related required information. Precision at its finest!

Plugins

Plugins are specialized components that execute planned actions. Each plugin is custom-built for specific tasks and system integrations, such as resetting passwords in Okta or adding accounts in Google Workspace. When combined, plugins unlock vast possibilities, automating processes seamlessly across your existing systems with Gaspar AI. They ensure efficiency without the need for reconfiguration, leveraging your current setup.

Clarifications

Clarifications are crucial when initial requests lack sufficient context for a solution. Our helpdesk actively seeks clarification to enhance vague requests. It engages in natural language conversations with employees to ask follow-up questions or searches integrated systems for additional helpful information. Instead of assuming or overlooking gaps, it uses two-way dialogue and system data to fill in missing information, gathering additional context to craft accurate responses.

Responses – Summaries

These are our virtual agent’s final responses. After processing retrieved information, including clarifications, system data, and actions taken, our helpdesk uses its generative capabilities to summarize crucial details in natural language and synthesize dynamic answers in a digestible format. Thanks to advanced NLP it generates accurate, helpful, human-like responses, ensuring that the generated answers are contextually appropriate, reliable and tailored to the employee's needs – not just chunky copied text.

Citations

When enabled by the customer on our platform, citations are essential additions to summaries, indicating the sources of summarized information, such as knowledge base articles or system data. They build trust and credibility by demonstrating that responses are based on authentic corporate information and facts. This verification instills confidence in employees regarding the accuracy of the summaries.

Clustering

Unanswered queries are not ignored; they are grouped and then translated into insights that uncover what’s blocking employee productivity. It then offers recommendations about high-impact automation and fresh content creation for the Knowledge Base . Gaspar AI thrives on turning challenges into opportunities!

Our Generative AI helpdesk building blocks

Now let’s see how we architected our Generative AI helpdesk to deliver best-in-class employee support. These are the core attributes that make Gaspar AI uniquely designed to meet each company’s demands:

  • Sector-specific high-quality datasets
  • State-of-the-art proprietary algorithms
  • Advanced LLMs fine-tuning
  • Fine-crafted prompts
  • Comprehensive evaluation process

Data: The heartbeat of Gaspar AI

At Gaspar AI, we understand that data is the lifeblood of any AI system. We adhere to five principles—relevance, consistency, uniformity, comprehensiveness, and anonymity. Our datasets, carefully curated and industry-standard, empower our AI with the knowledge it needs. We prioritize quality over quantity, ensuring our models deliver excellence. In case our models don’t have the expected performance for a particular customer, we create customer-specific models using training data mostly from that customer.

We do not use customer data to train our AI model

We're excited to confirm that we don’t use customer data to train our AI model. Although we have models tailored to individual customers using their specific data, our generic AI model relies on top-notch synthetic enterprise data.

This deliberate choice enhances platform security significantly. You can rest assured that your company’s data and sensitive information will remain private. Our AI model provides essential workplace intelligence, while customer data enhances and personalizes every customer’s model for an improved and tailored experience. Together, they create a helpdesk fine-tuned to each company without jeopardizing sensitive data. Our commitment to data security remains a priority, and this architecture reflects that dedication.

Algorithms: Where science meets art

Gaspar AI doesn't settle for mediocrity. We embrace the state-of-the-art (SOTA) in all our products, selecting algorithms that align perfectly with our problems and data. Our commitment to staying at the forefront of NLP developments means we're always pushing the boundaries.

When we find papers with methods that apply to our ML pipeline and are relevant to our problems, we conduct rigorous testing. It entails examining all aspects of the new method, including its appropriateness for our problems and data. Any new test goes through our home-bred evaluation pipeline.

Engineering: crafting brilliance

Behind the scenes, Gaspar AI's engineering competence shines. From fine-tuning large language models (LLMs) to perfecting Information Retrieval and GPT prompting, our engineers are the architects of an AI revolution.

  • LLM Fine-Tuning: Building on open-source LLMs requires expertise, and our engineers bring years of experience to the table. They navigate the complexities of NLP and Big Data, ensuring the best performance.
  • Information Retrieval: Querying our Information Retrieval system is an art. Our search engineers enhance keyword-based searches, correct spelling mistakes, and rank results for a seamless user experience.
  • GPT Prompt Engineering: The advent of ChatGPT was a game-changer, and Gaspar AI was quick to leverage its potential. Our engineers crafted prompts that deliver concise, human-friendly answers, elevating user experience to new heights.

Evaluation: The bedrock of excellence

We take evaluation seriously. It's not just a step in our process; it's a commitment to excellence. From testing new algorithms to maintaining dataset quality, our evaluation pipelines keep our AI systems sharp and effective, and our datasets up to standard.

Satisfying each company’s unique demands

By leveraging these building blocks, our generative AI helpdesk satisfies the unique demands of every company, helpdesk/IT agent, and employee:

  • Smooth collaboration: employees engage in interactive conversations with our AI virtual agent, just like talking to a live agent.
  • Human-like logic: our virtual agent understands user intent and context and selects the optimal solution while completely understanding your business systems.
  • Fact-checking: A meticulously refined evaluation model ensures answer accuracy by validating outputs with the sources used and assessing alignment with the original query.
  • Designed for scale: With robust ML operations and high customization capabilities, it is easy to build new use cases and extend to new languages while preserving critical business logic.

Gaspar AI: the generative AI helpdesk that shapes the future of work

To sum-up, the key elements fueling Gaspar AI's success are:

  • High-quality datasets: Carefully curated, industry-standard datasets drive our AI to excellence.
  • State-of-the-Art, proprietary algorithms: We don't follow trends; we use the best algorithms for our problems and data.
  • Comprehensive evaluation process: Rigorous evaluation ensures our AI systems stay at the pinnacle of performance.
  • Excellent engineering: From LLM tuning to Information Retrieval and GPT prompting, our engineering is a beacon of brilliance.

Together with an AI-powered, thorough process that identifies and plans the best solution to every employee query, Gaspar AI empowers helpdesk teams and employees alike to be more productive and enjoy happier support experiences.

This blog provides a glimpse of our ongoing commitment to innovation. We're dedicated to refining and enriching our functionality continuously. Gaspar AI is not just an AI solution; it's a commitment to shaping the future of AI-powered employee support. We're excited to embark on this journey with you, as we witness its immense potential unfold in the coming months and years. The future looks promising, and it's powered by Gaspar AI!

See how employees can use one unified Generative AI solution and enjoy smarter, automated support while freeing your helpdesk team’s valuable time. Schedule a personalized demo.