How we keep the decision with the operator
THE EDGE COMPANY builds artificial intelligence that supports the operator, who keeps the decision, and every figure in its reports traces back to a stored event.

Whoever runs wildlife control at an airport answers for what is done after an alert. The choice to act or to wait belongs to the person on duty, who also has to explain it later.
Any equipment added to that work has to respect the same line of responsibility. We design our work around that responsibility, from the site survey to the reports that reach the customer.
What does an alert ask of the person on duty?
An alert asks for a decision from someone who knows the airfield and the team's own procedures. It has to make sense to that person.
The artificial intelligence we build is human-centric: it is designed to support the operator, and autonomy is not its aim.
The detection itself is computer vision. Moving objects are detected and tracked, and specialized neural networks classify them from their appearance and movement. A classification made that way is still a model's judgment, and the person on duty is the one who weighs it.
A wildlife team may object that every new alert is one more thing to chase, on a shift that already has its own routine.
Part of the objection is true: any automatic detector makes mistakes, and we do not promise a system without them. An alert that nobody can check adds work and gives the person on duty nothing firm to decide on.
The person on duty therefore needs a way to check each alert. Anyone who reads the record later needs the same.
Why do we retrain the model on each site's own footage?
With BCMS® (Bird Control Monitoring System), each site starts on a pretrained model, which we then retrain on the site's own annotated footage. The retrained model has learned from the same place where its alerts will be read and checked.
A new site runs on the pretrained model until its own footage has been annotated and the model retrained. Anyone judging the first alerts at a new site should read them with that limit in mind.
The retraining is one step in our work at each site. We run the site survey, supervise the installation, commission the system, and provide remote support.
Who checks the detections?
Biologists in our data analysis team build species knowledge and validate samples of the detections. The team is led by one of them, who has worked on BCMS® since 2018.
Their work is a human check that sets a biologist's judgment beside what the models report. The validation works on samples, so a biologist does not see every detection.
The team's work has also been set out in public. The XXII Convegno Italiano di Ornitologia (Italian Ornithology Congress) was held in Lecce from September 8 to 12, 2025.
At the congress, Fabio Masci, who founded the company in 2017 and leads it, presented a talk on field results. Its abstract was written with the data analysis team.
Can a figure in a report be traced?
Every figure in a report traces back to a stored event. When an operator asks where a number came from, there is an event to point to.
Detection, classification, and storage run on a server with NVIDIA GPUs, normally at the site, with no cloud service. The server can also run on a closed network. The record a figure leads back to is kept on the same server that made the detection.
The person who answers for the decisions can then discuss a figure on what its event shows, with colleagues or with anyone who asks.
Reports carry the same limit as alerts, because each figure depends on the event behind it. Sample validation of the detections therefore matters for reports too.
Where does the decision sit when a deterrent is connected?
A deterrent is there to disperse birds from the area where they create a risk. Connected deterrents fire under the operator's rules, and the decision stays with the operator, as it does for an alert.
A wildlife team may still point out that, in the end, a connected deterrent fires. It does, and the rules it fires under are the operator's.
You can test this method on your own procedures. Choose one alert your team has handled and walk it through with our team, from its arrival to the decision that followed.
The walk-through shows, on a case you know, where a person checks and where the decision sits. To arrange it, write to us through the contacts page.
Technical documents
BCMS® is the product THE EDGE COMPANY builds today. Its capabilities and limits are described on bcms.ai, another site.
About THE EDGE COMPANY
THE EDGE COMPANY is an Italian deep-tech company, founded in Rimini in 2017 and led by Fabio Masci, one of its founders. It turns flight-safety experience into artificial intelligence for regulated, safety-critical environments: airports, wind farms, and industrial sites. Today it builds BCMS® (Bird Control Monitoring System), an AI optical detection platform.
Media contact
For interviews and information about THE EDGE COMPANY, write or call us.
+39 0541 670335
