What others have already done
Agentic AI in practice
What companies and public bodies are already doing concretely with agentic AI — in industry, transport, customer service and public services. Each case is explained in detail: what was put in place, how, and what to take away from it.
These cases were made public by the organisations named and are restated by Blue Lemon Agent by way of illustration. They are not our clients and we have no connection with them.
Food compliance
Food compliance: an AI assistant that answers, with sources
Trace One and delaware launched “Legi Food” in 2025, an AI assistant dedicated to food regulatory compliance: it answers complex regulatory questions while citing the applicable texts, in more than 50 languages. The publishers report up to 80% less time spent on regulatory monitoring.
The takeaway: Regulatory monitoring is a huge reservoir of time for AI — without taking the decision away from people.
Read the article →
Rome airports
Rome airports: an AI assistant to guide every traveller
The operator of Rome's airports (Aeroporti di Roma) launched an AI-powered traveller assistant in 2025, named ADRYX, available on WhatsApp and its website. Built on a multi-agent architecture, it informs passengers in real time (flights, gates, services) for a smoother journey.
The takeaway: An AI agent can unify the information across a complex customer journey: an ideal place to start.
Read the article →
Payments
Payments: AI serving reliability and fraud prevention
Payments specialist Nexi is deploying AI on two fronts: machine-learning fraud detection, in real time and at very large scale, and the operational reliability of its services, assisted by generative AI.
The takeaway: On fraud, an AI watches continuously what no team could cover by hand.
Read the article →
Customer service
Customer service: omnichannel AI supporting the teams
Riverty (financial services, Bertelsmann group) rebuilt its customer service around an AI-assisted omnichannel strategy, with Cluster Reply on Microsoft Dynamics 365. Phone, chat and email brought together across several markets and languages, for fast and consistent answers.
The takeaway: AI that is consistent across every channel frees your teams for the cases that matter.
Read the article →
Food & agriculture
Computer vision: speeding up varietal selection
Belgian manufacturer COSUCRA (ingredients from peas and chicory) automated the measurement of its chicory roots with delaware and the Digital Wallonia 4.AI programme, using computer vision — a key step in its varietal selection, previously done by hand.
The takeaway: Computer vision makes repetitive measurements reliable and frees up R&D.
Read the article →
Fraport
Fraport: computer vision smooths ground operations
At Frankfurt Airport, Fraport, Lufthansa and zeroG developed “seer”, a computer vision solution that automatically detects and timestamps the stages of aircraft ground handling (refuelling, baggage, boarding…), to improve punctuality.
The takeaway: Computer vision turns a video stream into operational decisions.
Read the article →
Audi
Audi: a multi-agent system to manage its cloud
With Storm Reply, on AWS, Audi deployed a multi-agent assistant named Devbot that supports its cloud teams across four areas: security, costs, infrastructure-as-code and architecture compliance. Several specialised AI agents work together.
The takeaway: Multi-agent systems also tackle highly profitable internal subjects, such as cloud management.
Read the article →
Industry
Industry: taking agentic AI to scale
Beyond demonstrators, AI agents are beginning to steer and optimise industrial processes by integrating with existing systems. Players such as Siemens presented AI agents connected to their automation tools in 2025; the challenge now is to industrialise measurable gains.
The takeaway: Agentic AI is no longer a proof of concept: plan for scale from the outset.
Read the article →
Data
Data: a modernised platform, boosted by GenAI
French group Lesaffre (yeast and fermentation) modernised its data platform with delaware: migration to SAP Datasphere, reporting through SAP Analytics Cloud and integration of generative AI, to make data access more reliable and simpler.
The takeaway: A sound, accessible database is the prerequisite for any useful AI.
Read the article →
Media
Media: subscription offers personalised by AI
Italian media group RCS MediaGroup (Corriere della Sera) implemented, with Go Reply on Google Cloud, an AI engine that personalises its digital subscription offers in real time according to each reader's profile — with better retention as the goal.
The takeaway: Personalising the offer at the right moment recovers margin left on the table.
Read the article →
Water
Water: detecting leaks before they become costly
Flemish water utility De Watergroep trialled AI leak detection with delaware: flow sensors, cross-referenced with the network's history, spot anomalies. Reported result: leaks detected up to two weeks earlier.
The takeaway: Detecting an anomaly early avoids the loss — and the bill — later.
Read the article →
Predictive maintenance
Predictive maintenance: an autonomous robot that inspects and anticipates
An autonomous quadruped robot (SPOT, by Boston Dynamics) combined with AI carries out inspections of structures and installations. Documented case: the Port of Hamburg authority evaluated it to inspect a large bridge, with a high-resolution camera and 3D scanning.
The takeaway: Anticipating a drift costs far less than suffering a shutdown.
Read the article →
FAQ
Frequently asked questions
What does agentic AI actually look like in practice?
Concrete results: unified customer journeys, maintenance anticipated, compliance accelerated, fraud better detected, offers better personalised. This page brings together 12 company cases, each explained in detail.
Are these cases only for large groups?
No. The same principles (an AI agent supporting a high-volume task, under supervision) apply to small and medium businesses and to the public sector, on a smaller scale.
Are the figures quoted guaranteed?
No: the results mentioned are reported by the companies or their providers, sometimes as a target or a pilot. We present them for guidance only. With you, we measure before we widen the scope.
And your first AI agent?
In 15 minutes, let's identify the most profitable one for your organisation.