One of the biggest challenges in effectively using data is getting the right insights to the right people in an efficient manner.
In most industries, decision-makers do not directly deal with data. Briefs are handed out, meetings are attended, and emails are sent—and these might happen before critical information reaches the people who need to act on it.
Aquant designed Service Co-Pilot to provide service leaders unrestricted data access and actionable insights. Over the past three years, Aquant developed Service Insights with the help of service leaders to assist decision-makers in comprehending their business, from technician performance and asset levels to customer insights and business-wide metrics.
Our recent work relies on something other than LLMs to better understand the data. Instead, it makes insights more accessible to users.
Presenting data analysis results is challenging because it requires telling a story that motivates business leaders to take action. It’s easier to do when the results relate to a single analysis and are straightforward. However, depending on the complexity of the analysis, it can take several hours or even days to deliver actionable insights.
We have been working on our Service Insights platform to improve service leaders’ abilities to tell a coherent story based on unknown results. This task is challenging, but we have been exploring, developing, and refining our process to generalize it better. We have started using large language models (LLMs) as part of this journey.
How it Started: Involving LLMs to Accelerate Product Development
Like many other data scientists, I got excited about Generative AI breakthroughs and started thinking about how it could impact existing models and product development.
The beginning of the project, alongside the emergence of ChatGPT, LLama, and Bard, was filled with uncertainty and excitement. It wasn’t entirely clear how this would impact our day-to-day, but one thing was for sure: it would be significant.
I still have screenshots of my coworker’s reactions from Zoom meetings where I showed results from the Proof of Concept. Since then, the team at Aquant has met and overcome the many seen—and unforeseen—challenges that came with adapting to new technology: token limits, load times, and the ambiguity of human language, to name a few. We have crossed the hurdle between an LLM product that could create value to a product that performs consistently.
From Insights to Value: Putting AI into Action
Although it is still early, we are witnessing great results in how our customers interact with their data. Our platform not only helps users access their data—it provides relevant insights and background information to extract meaningful narratives. We go beyond simple answers, forming narratives that provide a tailor-made, centralized view for any user question.
I am very proud of our ability to synthesize data in real-time to create a cohesive understanding of complex data sets—and we are continuously working to improve and refine this ability.
What You Can do With Aquant’s Co-Pilot for Service Insights
For example, let’s say a user wanted to determine which customer cost the most to service/maintain over the last quarter.
This question can be answered through a dashboard or SQL query, but the output will lack details. With Co-Pilot for Service Insights, we can get even more granular answers, including:
- The name of the relevant customer and their total cost.
- Plots that show how the customer compares to other high-cost accounts.
- Additional contextual information relating to customer costs, such as the average value over that time.
- Information regarding their risk profile—including any recent changes we may need to be aware of—if the customer is identified as a high-cost entity.
- Specific problematic assets, if applicable.
Most importantly, getting answers without translating data or researching context is easy. Service Co-Pilot is intuitive, speaks the service language, and corresponds to your daily business questions. This provides users with the information they need. It also aids in understanding the broader context of the results, allowing them to address any issues identified.
Taking Co-Pilot’s Insights to the Next Level
Co-Pilot for Service Insights has partial coverage of everything offered by the original product, Service Co-Pilot. Narrowing the scope allowed us to focus on the base methodology and logical backend.
With recent developments, we can offer consistent results to various questions. Up ahead, we are expanding Service Co-Pilot’s toolkit to allow increased customization, as well as more relevant and actionable replies. We are also constantly tweaking and optimizing to reduce wait times and improve accuracy.
Enabling Other Teams With Service Data Resources
Leveraging AI systems and generating actionable insights are the tip of the iceberg! AI can unlock various data types and resources for other service teams—including product manuals, video tutorials, and more.
At Aquant, we work closely with leading service companies to understand their main challenges in a fast-changing environment and how we can bridge these gaps by bringing technology and our service expertise.
If you are as passionate about AI technology as I am, you should check out Aquant in action.
I’m excited to uncover more details and truth in future blogs, but don’t hesitate to contact the team with any questions!
About the Author
Tommer Vardi, Data Scientist & Team Lead, Aquant
I am a Data Team Lead at Aquant, dedicated to leveraging Large Language Models (LLMs) to change how we interact with data. The team and I are driven by a passion for making data accessible and insightful. With nearly three years at the forefront of data science innovation at Aquant, I explore how LLMs can transform complex information into actionable knowledge.