How can you rethink the support request processing workflow to free up time for your support teams and improve the user experience? This is the challenge taken on by CPage, a French developer of administrative software for public hospitals, which made a bold decision: to make interaction with its AI chatbotmandatory as a first systematic filter before any support ticket can be created. During this webinar, Alexandre Letic, support manager at CPage, shares insights on the deployment of an AI chatbot using the Wikit Semantics platform: the methodology, quantified results, challenges encountered, and key success factors.
* For Gemini, the prompt is copied automatically: paste it into the input field to generate the summary.
Key Takeaways from This Webinar
- CPage, a developer of administrative software for public hospitals, has deployed an AI chatbot using the Wikit Semantics platform
- Usersare now required to go through the chatbotbefore creating a ticket, while the phone line remains available as a backup option
- -20% decrease in incoming ticketsstarting in the second month of production, with peaks of -30% to -35% in certain months
- Median resolution time reduced from 1.5 days to 1 day(-33%)
- The chatbotresolves up to 40 to 50 percent of support requests on its own, depending on the maturity of the knowledge base
- An average of 700 conversations per day,20% of whichtake place outside of business hours
- Organizing the knowledge base by business domainwas key to eliminating the issues caused by shared vocabulary across modules.
- AI hasnot reduced the support staff: the time saved is being reinvested in complex requests and documentation. The role has been revitalized, with virtually no turnover over the past two years.
- Key success factors: an up-to-date and reliable knowledge base, a critical testing phase, structured change management, measurable objectives set in advance, and a continuous improvement cycle.
The Context: Business Support Under Pressure
CPage handles approximately15,500 requests per year, of which 60 to 65 percent involve business or application support (with the remainder being incident-related). Users expect the same level of responsiveness, whether it’s a critical incident or a simple question about how to use the system. Three key challenges have emerged for the support team:
- Free upemployees' timeso they can handle high-value-added requests
- Maintain a high level of responsivenessto all requests, whether they are incidents or support requests
- Handle volume growth with the same headcount, without compromising customer service
Rather than a partial overhaul, CPage is starting from scratch in late 2024: consulting with 60 to 70 employees about what they like, dislike, and would expect from an ideal platform, followed by a request for proposals. Wikit stood out by presenting areal, working demo, rather than just a mockup. This was a deciding factor in choosing a partner.
Deployment: From the RFP to Production Launch
The project schedule was particularly tight:
- May 2025: Start of technical integrations
- December 1, 2025: Launch of the new platform
- January 31, 2026: End of the stabilization phase (operational stabilization was actually achieved in 5 days)
The chatbot draws on two data sources: the product documentation indexed in the Wikit Semantics console, and a FAQ database. The support team continuously updates this FAQ database whenever a question is asked for which there is no existing answer.
Use of the AI chatbot is now mandatory
A strong and deliberate organizational decision: CPage users must interact with the chatbot before they can create a ticket. To support this significant change, CPage has implemented several key measures:
- keeping its phone line open as a safety net,
- extensive outreach well in advance (special events, webinars, email campaigns) targeting both users and technicians,
- Assigning a full-time employee from the launch to monitor conversations in real time and assist users in formulating their questions.
Data Silos: A Prerequisite for AI Reliability
To succeed in this challenge, the chatbot still had to provide accurate responses. During the testing phase, CPage found that the chatbot generated “hallucinations” by confusing the terminology common to its three CPage modules (patient administrative management, economic and financial management, and HR). CPage therefore decided to segment the knowledge base by business domain so that each user could select their area of interest before asking a question. This testing phase served as a litmus test for the quality and relevance of the existing documentation. The preliminary data cleanup phase appears to be essential for any similar project.
AI for Business Support: Key Findings
- -20% drop in incoming ticketsstarting in the second month after launch, with peaks of -30% to -35% in some weeks
- The chatbotcan resolve up to 40 to 50 percent of support requests on its own, depending on the maturity of each institution’s knowledge base
- Median resolution time: decreased from 1.5 days to 1 day
- Annual projection: ~13,000 applications, down from 15,000 to 16,000 previously
- An average of 700 conversations per daysince December 2025 (5,000 users have access to support, 1,500 of whom are active each month)
- 20% of conversationstake place outside of business hours, highlighting the need for 24/7 availability
- Accelerated onboarding: A new support employee becomes self-sufficient in 6 months, compared to 9 months previously
Beyond the numbers, CPage observed an unexpected effect. In fact, its own developers and analysts began using the chatbot to verify their understanding of business processes, and many members—even those who were experts in the software—began using it to quickly generate internal procedures.
The Impact of AI on Support Teams
Contrary to a widespread fear, AI has not reduced the number of support staff at CPage. The time freed up for support technicians has been reinvested in providing more in-depth assistance with complex requests, producing documentation, and supporting the development teams. The result: an enhanced status for the role and virtually zero turnover within the support team over the past two years!
Key Success Factors for an AI Project Supporting Business Operations
- Define a clear functional requirementfrom the outset with a well-defined scope
- Set measurable and achievablegoals early on(the goal of a 20% reduction in incoming tickets, achieved by the second month)
- Maintain an up-to-date and reliable knowledge base and update it daily (up to 40–50 new FAQs per week during the first few weeks)
- Lead a structured change management process, both for customers and internal teams (communication strategy)
- Establish a cycle of continuous improvement: every unsatisfactory response is analyzed, escalated to a designated employee, and corrected at the source
- Have access to personalized , proactivesupport through weekly check-ins, both during the deployment phase and for continuous improvement.
