Periodic Reporting for period 2 - Lili.AI (Disruptive AI project management solution to drastically reduce the costs & risks of European large-scale construction projects)
Reporting period: 2023-09-01 to 2025-05-31
Introducing Lili.AI: Revolutionizing Project Management with Cutting-Edge AI
Lili.AI is a groundbreaking project management solution that leverages cutting-edge AI technology to help organizations successfully execute large-scale projects with budgets exceeding €100 million. These projects span various sectors, including infrastructure construction, engineering, waste management, and more, all of which play a pivotal role in promoting social and economic prosperity across Europe.
At its core, Lili.AI combines state-of-the-art AI models, pre-trained on real-world data from extensive project management endeavors, with award-winning natural language processing technology. This powerful combination enables Lili.AI to extract valuable insights from unstructured textual data sources such as emails, meeting minutes, correspondence, and progress reports.
One noteworthy application of our technology is its effectiveness in claims litigation. With the EIC accelerator grant, we are actively conducting trials of the Lili.Monitor technology which detects risks based on daily project textual data in real-time. This innovative approach addresses a longstanding challenge faced by organizations: the difficulty of detecting and addressing issues at an early stage. Delayed problem identification often results in the exacerbation of issues and substantial financial penalties, with the average claim amounting to approximately €54 million, according to Arcadis.
In summary, Lili.AI represents a cutting-edge technological solution that not only advances the field of project management but also promises to have a profound societal impact by helping organizations execute critical projects more efficiently, thereby contributing to the overall prosperity of Europe.
Lili.AI assist in robustifying our industrials margins by reducing exposure to claims: with better documented claims and also with early detection of weak signals.
Lili.AI is a groundbreaking project management solution that leverages cutting-edge AI technology to help organizations successfully execute large-scale projects with budgets exceeding €100 million. These projects span various sectors, including infrastructure construction, engineering, waste management, and more, all of which play a pivotal role in promoting social and economic prosperity across Europe.
At its core, Lili.AI combines state-of-the-art AI models, pre-trained on real-world data from extensive project management endeavors, with award-winning natural language processing technology. This powerful combination enables Lili.AI to extract valuable insights from unstructured textual data sources such as emails, meeting minutes, correspondence, and progress reports.
One noteworthy application of our technology is its effectiveness in claims litigation. With the EIC accelerator grant, we are actively conducting trials of the Lili.Monitor technology which detects risks based on daily project textual data in real-time. This innovative approach addresses a longstanding challenge faced by organizations: the difficulty of detecting and addressing issues at an early stage. Delayed problem identification often results in the exacerbation of issues and substantial financial penalties, with the average claim amounting to approximately €54 million, according to Arcadis.
In summary, Lili.AI represents a cutting-edge technological solution that not only advances the field of project management but also promises to have a profound societal impact by helping organizations execute critical projects more efficiently, thereby contributing to the overall prosperity of Europe.
Lili.AI assist in robustifying our industrials margins by reducing exposure to claims: with better documented claims and also with early detection of weak signals.
The project delivered a fully operational Big Data platform capable of ingesting, processing, and analyzing large volumes of unstructured documents in real time.
Activities performed included:
- Redesigning existing NLP and ML pipelines to enable distributed, parallel execution on large datasets.
- Setting up a scalable cloud-agnostic infrastructure.
- Developing and integrating connectors for enterprise systems especially Microsoft 365.
- Packaging the system for flexible deployment in client environments, either via portable Kubernetes clusters or multi-cloud setups.
Main achievements:
- Transformation of our initial processes into scalable real-time pipelines with full traceability and scalability.
- Deployment of ML/NLP models for document classification, entity and relation extraction at scale.
- Delivery of a production-ready system (TRL9) capable of supporting high-throughput semantic search, claim assistance, and dynamic risk monitoring.
The final outcome is a robust, multilingual, real-time AI platform, deployable across varied technical environments and capable of supporting future industrial use at scale. The models can be easily trained into another domain.
Activities performed included:
- Redesigning existing NLP and ML pipelines to enable distributed, parallel execution on large datasets.
- Setting up a scalable cloud-agnostic infrastructure.
- Developing and integrating connectors for enterprise systems especially Microsoft 365.
- Packaging the system for flexible deployment in client environments, either via portable Kubernetes clusters or multi-cloud setups.
Main achievements:
- Transformation of our initial processes into scalable real-time pipelines with full traceability and scalability.
- Deployment of ML/NLP models for document classification, entity and relation extraction at scale.
- Delivery of a production-ready system (TRL9) capable of supporting high-throughput semantic search, claim assistance, and dynamic risk monitoring.
The final outcome is a robust, multilingual, real-time AI platform, deployable across varied technical environments and capable of supporting future industrial use at scale. The models can be easily trained into another domain.
One of the hallmark achievements of our project is the transformation of real-time project document dataflows into machine learning-ready datasets through our robust ML pipeline. By efficiently converting complex project documents into structured datasets, and using active learning, we ensure that our AI continually refines its understanding and performance over time.
Also, this work led to the development of effective strategies for processing documents rich in domain-specific terminology and for extracting structured knowledge from data streams that reflect complex timelines and interdependencies.
Operating within industrial environments also surfaced essential non-technical challenges. We addressed the need to build trust with users and project teams by involving subject matter experts directly in the model training and correction process, using real, operational project data.
These outcomes mark key progress toward high-confidence, user-in-the-loop AI systems capable of supporting complex, document-intensive industrial projects.
Also, this work led to the development of effective strategies for processing documents rich in domain-specific terminology and for extracting structured knowledge from data streams that reflect complex timelines and interdependencies.
Operating within industrial environments also surfaced essential non-technical challenges. We addressed the need to build trust with users and project teams by involving subject matter experts directly in the model training and correction process, using real, operational project data.
These outcomes mark key progress toward high-confidence, user-in-the-loop AI systems capable of supporting complex, document-intensive industrial projects.