The
project client is one of the largest fast-food chains in Russia, with up to
1,000 locations nationwide.
The company's Service Desk receives a large number of inquiries from restaurant employees (cashiers, administrators) regarding non-functioning equipment: cash registers, monitors, scanners, payment terminals, and other issues.A significant portion of these inquiries also relate to application failures and various technical incidents, which further increased the workload of the first-line support team. At the same time, the company experienced a number of difficulties in operating the service for the following reasons:
- The setup and troubleshooting instructions contained numerous technical details and were not simplified to simple steps. This made diagnostics time-consuming, and restaurant equipment downtime was prolonged.
- The system had accumulated over a hundred documents with detailed instructions, which were difficult to categorize and quickly locate upon request. Finding the right document when an employee contacted it was extremely difficult.
- It took new employees up to three months to master the entire knowledge base and begin providing high-quality and confident answers to users. This significantly slowed down newcomers' ability to work independently and increased the workload for experienced colleagues.
- All complex requests were escalated to experienced specialists, which reduced the overall effectiveness of support and led to customer dissatisfaction. This increased incident resolution times and created reputational risks for the customer's business.
Key Challenges
- Implement an AI agent that would help service desk staff find the necessary instructions in real time.
- Ensure fast loading and indexing of the customer's entire existing knowledge base.
- Configure a specialized glossary and prompts to account for industry terminology so that the AI agent correctly recognizes equipment names and the customer's internal designations.
- Integrate the AI agent with the existing Service Desk system for seamless use in the workflow, eliminating the need for additional systems and switching between interfaces for staff.
- Confirm measurable economic benefits (reduced solution search time, reduced number of escalations) in a pilot project as a basis for replicating the solution.
Implemented our solution
- The solution is based on ICL Services' own Pro.GenII platform, which combines advanced large-scale language models (LLM) and rapid acquisition and generation (RAG) technology. This approach allows the AI agent to respond not abstractly, but strictly based on relevant documents and customer regulations.
The product was created to automate ticket processing and intelligently search knowledge bases and company files. It integrates with ITSM systems, provides document recognition, and extracts data from attached files, reducing operator workload by a factor of five and processing up to 80% of tickets without human intervention. These capabilities determined the choice of Pro.GenII for the pilot project—the platform was originally designed to work with large and diverse internal knowledge bases.
The project was a pilot implementation to demonstrate the technology's potential within the customer's environment.
Platform Preparation and Adaptation
ICL Services specialists deployed access to a corporate instance of the Pro.GenII solution for the Service Desk support team. System setup took approximately one hour:
- uploading the entire existing knowledge base (53 documents with instructions and regulations),
- creating a glossary to correctly handle specific terms (equipment names, internal corporate designations),
- setting up a system prompt: the system was positioned as "Expert in Technical Support for Restaurant Equipment."
Integration and Training
The team configured seamless integration with the Service Desk system via open APIs. Since the solution is based on the familiar GPT chat format, support staff underwent extensive training on how to use the AI interface. The familiar chat format reduced the time it took staff to adapt to the new tool.
Industrial Operation in a Pilot Environment
A process was established whereby a service desk employee entered each new user request into the AI agent in real time. This allowed the solution to be tested not in a pilot environment, but on a real stream of requests from operating restaurants.
The system instantly processed the request (for example, "The monitor in the service area is not working") and provided step-by-step troubleshooting instructions (check the cable, reboot the monitor, reset the IP address, and restart the server if necessary).
If necessary, the employee could also request a detailed document for an in-depth investigation of the problem.
Results
- Time to find a solution has been reduced by 70%: employees no longer waste time manually searching through multi-page documents. Operators began to receive a ready answer in seconds instead of tens of minutes.
- The workload of experienced specialists was halved: thanks to high-quality AI tips, beginners were able to independently process complex requests without escalation. Experienced specialists were able to switch from routine issues to truly unusual incidents.
- The quality of responses increased by 80+%: they became more structured, accurate and complied with current regulations. This directly reduced the number of repeat calls and complaints from restaurant employees.
- The onboarding process for new employees has accelerated. The time it takes for newcomers to start working independently has been reduced from 3 months to several weeks.
- The pilot successfully confirmed the technological effectiveness of the Pro.GenII solution for complex Service Desk systems with a large volume of technical documentation.