How Silae nearly doubled support deflection with a conversational AI agent
Company Overview
Silae is the French payroll platform trusted by 6,000 expert partners, from accounting firms to HR integrators, to manage pay for 8 million employees every month. Built on a single engine covering payroll, DSN, workforce management and the employee portal, Silae handles over 900 collective agreements and continuous regulatory updates, so partners can focus on deploying and supporting their clients rather than chasing compliance.
Challenge
eDocPerso generates the largest volume of support requests in the Silae ecosystem. eDocPerso is the employee-facing portal where 8 million workers access their payslips every month. At that scale, even a modest percentage of users needing support generates an overwhelming ticket volume, which is why deflection is a critical metric for Silae. The previous setup relied on a Zendesk bot. Deflection sat around 40%, so most questions still landed on the support team, and the field kept sending signals that the experience was not good enough.
What struck us most was the speed of deployment. From our first meeting to a product running in production, AGO moved fast while taking our specific constraints into account from the start. They adapted to our context and built a solution that genuinely fit our needs.
Solution
Silae partnered with AGO to replace the old bot with a conversational AI agent grounded in their own knowledge and data from internal APIs. The agent answers in natural language and holds a real conversation, adapting to each request. AGO took Silae's specific constraints into account from the start and personalized the solution to fit their context. The collaboration stayed agile throughout, with fast iterations on feedback.
Result
Deflection rose from around 40% to between 75 and 85%, which means a large share of tickets never reach the support team. The experience improved a lot, and the dissatisfaction signals the team used to get from the field disappeared. Silae now relies on AGO's CX evaluation to monitor conversation quality. The results convinced the team to position AGO as the first option across customer service and to extend its scope to other products.
Favorite features
Some of my favorite parts are things we never even asked for. The CX evaluation system scores the quality of every conversation, the dashboards keep getting sharper, and the MCP server and APIs let us connect the agent to the wider Silae ecosystem.
CX evaluation system to score conversation quality
Refined analytics dashboards
MCP server and APIs to connect the agent to the wider Silae ecosystem
A sovereign stack for sensitive HR data
Payroll and HR data is highly sensitive, so the agent runs on a stack built for European sovereignty and compliance from the ground up.
European AI model
Powered by Mistral, a European frontier model provider.
Hosted in Europe
Data and servers stay within the European Union.
SOC 2 and GDPR
SOC 2 and GDPR compliant.
Ready to reimagine your customer experience?
Deploy AI agents that truly understand your customers and your business. Book a demo with our team to see Ago in action.