How a Catering Company Revolutionized Welfare App Customer Care with an AI Voice Agent
Discover how a company with 11,000 employees automated support for its welfare app with an AI agent: increased resolutions and autonomously managed tickets.
An AI welfare customer care voice agent enabled a Catering Company with 11,000 employees to automate the first level of support for its welfare app. In eight weeks, the assistant increased first-contact resolutions and autonomously opened complete tickets when necessary. The result: more stable employee app support, predictable timelines, and a human team focused on complex cases. ## TL;DR - +33% goal met in two months: more problems resolved at first contact, without escalation. - 100% automation in app access ticket creation for unresolved cases. - +62% in average duration: more complete and resolution-oriented conversations with the AI voice. - Scalable 24/7 employee app support: less pressure on the internal contact center. - Integration with ticketing and HRIS: complete data for the second level.  ## The Problem: A Contact Center Overwhelmed by Repetitive Requests Before implementation, the internal contact center manually handled high and homogeneous volumes of requests: credential recovery, login errors, navigation difficulties. There was no first-level automation and no structured automated app access ticket flow. Scaling during peak hours was impossible without increasing staff, directly impacting perceived quality and welfare app adoption. ### Key Issues Identified - Calls concentrated on recurring, low-value problems. - Operators engaged in repetitive tasks, with little time for complex cases. - Tickets opened manually, incomplete data, and variable processing times. - Unscalable service dependent on team availability. ## The Solution: An AI Voice Agent Integrated into Customer Care DeepAgent implemented an AI voice agent trained on the context and workflows of the welfare app. The agent manages the first level: autonomous resolution of frequent cases; for cases not resolvable at first contact, automatic ticket creation with all collected information. ### How the Operational Flow Works 1. Employee identification and problem classification. 2. Initiation of the resolution path for access, credential reset, and basic navigation. 3. If escalation is needed, the agent automatically completes and submits a comprehensive ticket. 4. The human contact center becomes a second-level point for exceptions.  ## Measurable Results: Data from Two Consecutive Periods The impact is visible on all key KPIs and, importantly, shows a month-over-month improvement trajectory. | KPI | Period A (Jan 19–Feb 19) | Period B (Feb 19–Mar 17) | Delta | |---|---:|---:|---:| | Total Calls | 148 | 134 | -9.5% | | Completed Responses | 147 | — | — | | Goals met (problems resolved) | 66 | 88 | **+33%** (+22) | | Goal rate | 44.9% | 65.7% | +20.8 pt | | Average call duration | ~2.52 min | ~4.09 min | **+62%** | | Minutes consumed | 373.0 | 547.5 | **+46.8%** | ### What This Improvement Means - More than six out of ten calls successfully closed at first contact in Period B. - The longer average duration indicates more in-depth dialogues: step-by-step instructions, outcome verification, complete tickets. - 100% automation in creating unresolved tickets: no lost data, rapid transition to the second level.  ## Why Voice AI Works in Welfare and HR - Predictable requests: Access issues follow recurring patterns, ideal for automation. - Consistent quality: Employee app support does not depend on human team shifts. - Structured data: Voice AI for employee issue management transforms every call into useful information for follow-up. ## Lessons Learned and Replicable Best Practices - First map request types to train the agent on high-volume cases. - Evaluate average duration as a qualitative, not just quantitative, metric. - Design escalation and app access ticket automation before go-live (fields, routing, SLAs). - Monitor the goal rate as the primary KPI for the effectiveness of automated support.  ## Extension: From Welfare App to Distance Learning and Company Canteens Given the results, the company is extending the approach to two new areas: distance learning (FAD) and canteen management via a dedicated app. The same principle applies: an AI voice agent for each context, integration with existing ticketing, and continuous monitoring of the goal rate. ## Discover How DeepAgent Can Do the Same for Your Company DeepAgent's AI voice agent has made employee app support scalable: more first-contact resolutions, complete tickets, and a human team focused on value. Do you want to measure the impact in your context? Request a free demo: you will be called back at +39 068 384 5191. ## Frequently Asked Questions ### How long does it take to implement an AI voice agent for welfare customer care? On average, 3–6 weeks. It starts with discovery and mapping of use cases (access, credentials, navigation), then agent configuration, ticketing integration, and a controlled pilot. Progressive go-live allows optimizing the goal rate and app access ticket automation without interrupting employee app support. ### How does it integrate with existing systems (ticketing, HRIS, welfare app)? The agent connects via API to ticketing systems (e.g., ServiceNow, Jira Service Management) and HRIS/SIAM for identity verification. Data collected during the call automatically populates ticket fields. Integration ensures 100% automation in app access ticket creation and a searchable history for the second level. ### Which KPIs should I monitor to evaluate effectiveness? The main ones are: goal rate (first-contact resolutions), number of goals met, average call duration, escalation rate, total minutes, internal CSAT. In the analyzed case: **+33%** goals met, goal rate from **44.9%** to **65.7%**, **+62%** average duration, and **100%** ticket automation, all indicators of effective AI voice employee problem management. ### How is employee privacy and data security managed? Minimization and encryption principles are applied in transit and at rest. The agent accesses only the data necessary for employee app support, with an audit trail on every call. Integration with Identity and ticketing respects roles and permissions; logs are pseudonymized, and retention policies comply with GDPR. ### What happens in complex cases requiring human intervention? The agent escalates to the second level by creating a complete ticket with context, steps already taken, and partial outcomes. The human operator picks up from the exact point, reducing time and repetitions. This hybrid model maximizes the goal rate and satisfaction, leaving recurring volume to AI and exceptions to people.