Case study ·
Sekoia pushes its agentic SOC: what SPIE found when testing it
Original title, in French: “Sekoia pousse son SOC agentique : ce que SPIE a constaté en le testant”
- Publication
- Silicon.fr (France)
- Author
- Clément Bohic
- Published on
- 15 October 2026 · read in the page structured data · updated on 15 October 2026
Summary
The SPIE group, the first tester of Sekoia's Elevate agentic SOC module, presented its feedback at the Assises de la sécurité. Deployed since July 2026 to handle part of SPIE's 43,000 monthly alerts, the Roy agent showed shorter research times and accurate analyses, according to Sekoia CTPO Georges Bossert. Of around 10,000 alerts handled in September, 96% of verdicts were not changed by analysts and SOC teams became 2 to 5 times faster, although true positive detection remains imperfect (2 confirmed out of 411 predicted). SPIE's SecOps manager Dimitri Farcy and group CISO Sébastien Louis describe a transformation of level 1, refocused on coverage, with no job cuts, and upcoming discussions with their MSSP on sharing the value created.
Summary based on the full text of the article.
Read the article on Silicon.fr →Figures cited
-
43,000 alerts/month
Alerts received per month across SPIE's global scope
SPIE global scope
« Sur son périmètre global, SPIE reçoit environ 43 000 alertes par mois. »
-
96 %
Share of verdicts left unchanged by SPIE analysts
SPIE, around 10,000 alerts handled in September 2026, 11 analysts consulted
« Il avance deux indicateurs en particulier : 96 % des verdicts inchangés par les analystes**, et des équipes SOC « entre 2 et 5 fois plus rapides ». »
-
— factor (x2 to x5)
Speed gain for SPIE's SOC teams
SPIE
« des équipes SOC « entre 2 et 5 fois plus rapides » »
-
35,000 alerts
Alerts covered by the feedback presented at the Assises de la sécurité
SPIE, Elevate test
« Le retour d'expérience présenté aux Assises de la sécurité couvre un peu plus de 35 000 alertes. »
-
2 out of 411 predicted
True positives confirmed among Elevate's predictions
SPIE, September 2026, around 10,000 alerts
« Elevate s'est avéré moins bon sur les vrais positifs : des 411 prédits, seuls 2 en étaient. »
-
94.4 %
Elevate's weighted accuracy on a subset of alerts
SPIE, around 5,000 alerts, about fifteen rules
« La précision pondérée atteint 94,4 %. »
-
1,000 customers
Number of customers on which Elevate was deployed before general availability
Sekoia, before GA at the end of September 2026
« Avant d'annoncer la GA d'Elevate, Sekoia l'a déployé chez 1000 clients. »
-
88 seconds
Average Elevate investigation time among Sekoia customers
Installed base of 1,000 Sekoia customers, before GA
« Il annonce un temps moyen d'investigation de 88 secondes, pour 90 % de verdicts non modifiés. »
-
90 %
Share of unmodified verdicts across the Sekoia installed base
Installed base of 1,000 Sekoia customers, before GA
« Il annonce un temps moyen d'investigation de 88 secondes, pour 90 % de verdicts non modifiés. »
Classification
- Country
- France
- Company size
- Large companies
- Industries
- Software vendors, IT services companies, telecoms · Energy, utilities, environment
- Companies
- Spie (deploys AI), SPIE ICS (cited), Sekoia (supplier)
- Keywords
- cybersecurity, agentic soc, ai agents, security operations, lessons learned, threat detection
Entry written by Marsanne Studio from the cited article, with the help of an AI model (claude-sonnet-5). The rights to the article and its illustrations belong to Silicon.fr and its author. Entry translated from French by Claude Sonnet 5.5.