.svg)
During its first three months of operation with our clients, the Helpy AI agent resolved 60% of Level 1 tickets without any human intervention, with an average resolution time of 3 minutes. Period measured: June, July, and August 2026.
These figures are published because they are measured, not estimated. The French IT outsourcing market talks a lot about support automation but very little about its actual results. Here are ours, along with the methodology used and their limitations.
The Numbers
60% of Level 1 tickets were closed without human intervention, on average over the period
Average resolution time: 3 minutes
1,889 Level 1 tickets processed over three months, or an average of 630 per month
55 clients under full-service management contracts, all of whom have been migrated to Helpy since June 1, 2026
24/7 availability, including nights and weekends
Month-by-month breakdown: 680 tickets in June, 720 in July, 489 in August.
The Method
Scope. The 55 clients under full-service management contracts, all of whom have been using Helpy since June 1, 2026. This is not a selected sample, nor is it a projection: it represents the entire customer base under contract during this period.
Period: June, July, and August 2026—three consecutive months. The 60% rate is an average for the entire period, not a peak.
Counting rule. A ticket is counted as resolved independently only if no technician has worked on it. Any ticket handled by a human—even if only partially processed by the agent—is excluded from the rate.
Source: The Helpy monitoring tool, which tracks each ticket and how it was resolved.
What These Numbers Don't Tell Us
A barometer is only useful if its limitations are stated. Here are ours.
The number of tickets in August fell to 489, a 32% decrease from July. This is the summer lull, and we are reporting the figures as is rather than smoothing the curve over the quarter. A barometer that shows fluctuations is considered reliable; a flat barometer is called into question.
Three months is a short time. This survey is a starting point, not a trend. The next edition will cover September through December 2026, using the same scope and methodology, so that the two surveys will be comparable.
Le taux progresse avec la maturité de la base de connaissances. Il repose ici sur une exploitation menée depuis 2010 et sur des parcs déjà outillés. Un environnement moins documenté obtiendra probablement moins, au moins les premiers mois.
Why We Publish These Figures
An automatic resolution rate is touted everywhere but rarely substantiated. Publishing the scope, the time period, the counting rule, and the month-by-month breakdown allows you to assess the value of the figure rather than simply taking it at face value.
We built Helpy for ourselves before offering it to our customers. Releasing it without any parameters would have been just another sales pitch. Releasing it along with its methodology and limitations makes it a verifiable benchmark.
Further information
The deployment process, the agent's architecture, and the challenges we encountered are detailed in our post-deployment review on Helpy.
Regarding the solution itself: Helpy, our AI-powered help desk
About this survey
Data from the Helpy monitoring tool, June through August 2026, covering the 55 clients under full-service management contracts. Published in September 2026. Contact: s.amara@it-systemes.fr
IT Systèmes, ESN française fondée en 2010, quatre implantations : Paris (Malakoff), Lyon, Annecy et Bordeaux. Certifié ISO 27001 pour nos activités de conseil, projets, sécurité et infogérance (AFNOR Certification), Microsoft Solutions Partner.
.svg)