We continue to discuss how AI was integrated into Gurtam’s operations. In a recent article, we explored how the flespi team uses AI to enhance technical support and manage GPS tracking devices. Now we're chatting with Artem Filimonchik, the Head of Technical Care L2 team at Wialon, who has headed the implementation of AI into his team’s operations, with the solutions launching recently.
Artem Filimonchik, the Head of Technical Care (L2) at Wialon, Gurtam
AI to assist technical support teams
- Artem, could you share how AI has become integral to your role? What specific needs does it address in your work?
Over the last year, AI has become a buzzword, prompting us at Gurtam to explore how we could use it internally. This exploration began with the AI Hackathon held at Gurtam, which offered a deep dive into AI’s capabilities. They became our passion, and we (me and Natalia Evpak, the Head of Technical Care (L1) team) decided to use them in our teams' daily routine work.
Subsequently, I pursued online courses to develop our first AI tools. We initiated with an internally used AI-powered bot that suggests GPS tracking hardware based on user queries. This project was foundational for what we now have. During the Hackathon, we focused on the server-based Wialon version, which required a solution that could provide precise outcomes. We started with a more contained solution, anticipating significant user benefits like immediate AI responses during server issues reducing wait times for technical support.
Artem Filimonchik, Natalia Evpak and their team at the internal Gurtam AI Hackathon, February 2024
We also leverage AI to simplify our daily routines, particularly in training and education. Platforms like ChatGPT and Copilot are invaluable, helping non-developer technical care staff gain knowledge in new technology fields and also automate daily routines.
- Is ChatGPT your primary AI tool, or do you utilize various AI technologies?
Yes, ChatGPT is a primary tool, but we explore other technologies as needed. For instance, I use Google’s Gemini for tasks requiring precise charting because it integrates well with Google Sheets, saving time spent on my managerial tasks. AI has fundamentally altered how I access and analyze information — I simply can’t remember when I last Googled anything.
We also employ AI for translations and business communications across different languages, ensuring our messages are polished and professional.
- Can you delve deeper into the AI-driven solutions you've implemented to enhance process efficiency? Is the AI tool that suggests hardware configurations intended solely for internal use?
This tool primarily aids our project implementation team by providing tailored hardware recommendations crucial for deploying effective fleet management solutions for Wialon partners. Our extensive experience with diverse telematics devices has been digitized and formatted for AI use, bringing the equipment selection process to a new level.
Our overarching goal for AI in hardware management is to provide setup assistance across the broad array of GPS tracking devices integrated with Wialon, currently over 3,500. Collaborating closely with the flespi team, who have excelled in AI-powered device management, we aim to extend these capabilities directly to our clients to enhance their operational efficiency.
- Looking ahead, what are your aspirations for further developing this AI initiative?
Our vision involves three specialized AI assistants: one for managing hardware-related tasks, another for the server-based Wialon administration, and a third for the Wialon API. These projects are currently underway, and we're continuously learning and adapting. Automating technical support ticket handling remains a challenge, requiring the integration of diverse information sources into a sophisticated, AI-friendly knowledge base.
Creating an AI assistant has involved less programming than we anticipated — about 20% of our efforts — with the remaining time spent on AI training, developing guidelines and revising our databases. We've restructured the knowledge base multiple times to suit AI requirements, eventually finding a format that delivers precise responses. This is an ongoing task as we continuously adapt to new client inquiries.
- How many team members are actively involved in developing these AI solutions?
Seven of my 11-person team are engaged in AI development. My like-minded person, Natalia Evpak, covers much of inspirational development. We’ve noticed that balancing this initiative with client-driven priorities remains a challenge, as client requests must always come first. However, incorporating AI into our operations has reinvigorated the team, making the activity meaningful for the company. It's common knowledge that technical support staff sometimes lose motivation; in our case, implementing AI has helped to balance routine.
- What has been the most significant challenge you've faced during the past six months of development?
It surprised me that despite the widespread interest in AI, the information available could be superficial. Diving deeper into specialized topics requires significant effort, as the field is still evolving. It’s akin to pioneering; while the concept of an AI bot for technical support seems on the surface, the implementation is complex. We at Gurtam are lucky to have learned from our experiences and benefitted from the knowledge we share across different teams within the company.
Aliaksei Shchurko, CEO at Gurtam, takes a leading role in the company with AI tools
- Have there been any surprising results or outcomes from your AI projects?
One of our most successful AI applications was developed in just 40 minutes last spring, dramatically saving time on project documentation. We typically handle extensive lists of questionnaires related to system safety — a critical aspect of technical documentation for tenders. A technical support specialist would usually spend considerable time deciphering terminology, comprehending legal terms, and sourcing relevant information. Our legal team consolidated recent questionnaire data into a robust knowledge base, which we optimized for AI. This adaptation has proven exceptionally efficient, especially in handling the formal language typical of such documents. Previously, processing an average questionnaire took an entire workday; now, the AI bot completes it instantaneously.
It’s important to note that all information fed to the chatbot is securely reviewed internally at different levels. We construct the knowledge base ourselves, ensuring no sensitive data is ever exposed or shared.
- What are the complexities of processing source data to make it suitable for AI applications?
A major misunderstanding was assuming we could simply anonymize a client’s ticket, upload it to the AI, and expect an accurate response. The reality is quite different. Client inquiries can vary significantly, even on the same topic. It’s our responsibility to train the AI bot to recognize these variations and deliver precise responses. We've developed an optimal data storage format to facilitate accurate AI processing. Nevertheless, it’s an ongoing process, and I anticipate we will update our approach within the next three months.
Initially, I thought training AI would be similar to training a new team member — intelligent but needing guidance. However, the actual process differed drastically: imparting knowledge to a human is straightforward, whereas instructing a robot requires precision in how information is presented.
- What kind of feedback do you hope to receive from your clients regarding the AI tools?
From my observations, people are divided into two groups: those who embrace AI and tolerate its imperfections and those who are skeptical and sometimes pleased when it fails. As AI becomes more prevalent across various sectors, I expect the first group to grow, with people becoming more receptive to new technologies.
I emphasize to my team that our primary goal with AI is to enhance client experience, not to reduce costs. At Wialon, we’ve established a high standard for technical support, as confirmed by our client surveys, and we are committed to maintaining, if not exceeding, these expectations. AI bots significantly contribute by providing prompt, reliable responses based on a well-maintained knowledge base.
- In your ideal world, how will AI and humans cooperate?
We are developing an AI-powered support agent that operates across all Wialon spheres. It can sort incoming requests and handle them based on their specifics, referring some instances to human oversight. This agent, integrating multiple specialized bots, relies on a variety of analytical tools to process inquiries. By year's end, I aim to have a system that quickly handles basic inquiries.
Our roles are evolving. We’re shifting from mundane tasks to focus on training AI and tackling more complex issues that bots cannot manage. These intricate tasks will often require preliminary AI work, such as research and log reviews, but the final decision-making for complex queries will remain a human responsibility.
- Do you perceive any risk of AI technologies replacing human roles within technical support teams?
I don’t foresee AI replacing our team, especially since Wialon partners handle initial support levels, and my team addresses more complex challenges. However, there is substantial potential for AI to enhance support capabilities at the partner level. Many telematics service providers operate with limited staff; by adopting AI, they could drastically reduce overhead, needing only a few specialists to manage and train AI systems, while bots handle the bulk of first-level support.
- What advice would you give other professionals considering integrating AI into their business processes?
-
-
-
Start with the fundamentals. Understanding how the technology functions will set realistic expectations and guide practical application.
-
Prepare for extensive testing and experimentation — AI is a powerful tool but not a panacea. Reflecting on the progress made from our initial bot capabilities to now, it’s clear that early adoption facilitates greater advancements.
-
Above all, prioritize customer experience. Always consider whether you would be satisfied with the AI’s response if you were the client. While we cannot foresee every scenario, we strive to ensure our clients are happy with the support they receive.
-
-
Recommended courses and resources
Artem has explored several resources to deepen his understanding of AI, particularly in the context of technical support. Below is his list of recommendations for anyone interested in implementing AI in their technical support role:
Prompt Engineering Resources:
-
Zero Shots and Few Shots Approaches: Start by understanding these foundational techniques to enhance the performance of AI bots.
-
Gemini for Google Workspace: A comprehensive guide that is applicable to various AI models.
-
Additional Prompt Engineering Information:
LinkedIn Learning Courses:
Documentation from Leading AI Developers:
These resources have been instrumental in Artem's AI journey. They offer both theoretical insights and practical applications for enhancing AI integration within technical roles.