It support · Production

Geodis builds an AI IT Support Agent in Dataiku to cut ticket assignment time by 60% and save 30 minutes per ticket

The problem

Geodis's global IT service desk suffered from manual triage and misclassification that drove unnecessary escalations, while valuable resolution knowledge scattered across past tickets and documentation remained hard to find and reuse consistently.

Workflow diagram · grounded in source
1
Incoming ticket received
trigger
“For each incoming ticket, the agent:”
2
Classify issue and predict routing
ai_action
“Classifies the issue and predicts the correct support level and resolution group”
3
Identify similar past tickets
ai_action
“Identifies similar historical tickets”
4
Suggest resolution steps
ai_action
“Suggests resolution steps using internal technical documentation and past cases”
5
Human review in ServiceNow
human_review
“All recommendations appear inside the ServiceNow interface, where members of the support team can review and apply them with a single click. The AI agent augments human expertise rather than replacing it, empowering support personnel to …”
Reported outcome

Embedding an AI agent in ServiceNow delivered a 60% reduction in ticket assignment time and saved IT teams around 30 minutes per ticket, with tickets routed more accurately, unnecessary escalations reduced, and Level 1 teams resolving a higher share of issues.

Reported metrics
Ticket assignment time60%
Time saved per ticketaround 30 minutes per ticket on average
Reported stack
DataikuServiceNow
Source
https://www.dataiku.com/stories/detail/geodis/
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Embedding an AI agent in ServiceNow delivered a 60% reduction in ticket assignment time and saved IT teams around 30 minutes per ticket, with tickets routed more accurately, unnecessary escalations reduced, and Level…

What tools did this team use?

Dataiku, ServiceNow.

What results were reported?

Ticket assignment time: 60%; Time saved per ticket: around 30 minutes per ticket on average (source-reported, not independently verified).

How is this it support AI workflow structured?

Incoming ticket received → Classify issue and predict routing → Identify similar past tickets → Suggest resolution steps → Human review in ServiceNow.

Related it support cases