Matilda is a robot trained with Machine Learning techniques that automates the reading and coding of procedural notifications. Matilda collects the notification, “reads” it and “extracts” its most relevant conclusions (procedural milestones, appointments, amounts, participants, deadlines) and presents them in a format adapted to the needs of each client, so that they can take advantage of this information to feed their procedural history or even trigger automated work sequences.

How it works

We determine the type of procedure whose notifications will make up the main body of the Project (Verbal, Ordinary, Executions, Monitors…)

We determine which entities, intentions and milestones make up the information that our client will consume: NIG, Procedure, Amounts, Dates, Deadlines and exit milestones (judgement, decree of admission of claim, appointment of preliminary hearing, order to evacuate, etc.)

We determine where the notifications will be collected from (mailbox, repository, lexNET, etc.), their structure and the format in which the results will be presented (json, Excel) and how these results will be integrated (annotation in the procedural agenda, in an intermediate database, etc.)

A presentation is made of the universe of notifications that make up the Project and the veracity thresholds, exception criteria and output voices are determined

Different pilots are carried out with real information that is then verified by the client’s team

Matilda is integrated gradually, first with a mixed model where a notifier validates the results, to move in the last stage of the project to an autonomous model


80% of notifications processed automatically

Accuracy rates of 99.99%

In a mixed model (one notifier validates the output provided by Matilda) increases in throughput by 700% (one notifier goes from 150 notifications per day to 750)

In a stand-alone model 80% reduction in workload

Identification of a common denominator allowing the linking of a notification to its original dossier in 98% of cases

More than 200 exit milestones

More than 20 types of parametrised cases

More than 30.000 notifications successfully processed per day