Document Processing
Intesa Sanpaolo's supervisory strategy team faced information overload from the massive and fast-growing volume of banking supervision publications, best practices, statements, interviews, and studies, making manual monitoring and trend analysis difficult without technological support.
Reported impact
The tool helped Intesa Sanpaolo process large volumes of supervisory information more efficiently, gave experts a broader strategic view of trends and developments, and delivered results strong enough to earn external recognition through the John McCarthy "Manager + AI Success Stories" Award.
Intesa Sanpaolo built and deployed Lisa (Linguistic Intelligence for Supervisory Awareness), an internal machine-learning and NLP tool for its Group Supervisory Strategic Steering Department, with support from the Chief Data, AI, Innovation and Technology Area and PwC. Lisa ingests large volumes of banking supervision publications from selected external sources including institutions, think-tanks, and consultancy firms. It uses natural language processing algorithms to read text at far higher speed than humans, analyze content for patterns of meaning, and search objectively for correlations across documents. The system then surfaces emerging supervisory themes and predictive signals so a dedicated internal team can validate results, refine the model's focus, and expand coverage to future-relevant issues. The outputs give experts a broader, more strategic view of supervisory developments and potential risk connections such as climate change and credit risk.
Banking supervision publications from selected external sources such as institutions, think-tanks, and consultancy firms are fed into Lisa, where NLP algorithms read and analyze the texts for patterns of meaning and correlations; the resulting insights are reviewed by a dedicated team and used by Intesa Sanpaolo experts to monitor supervisory developments and future trends.
"Human intelligence plays a decisive role in interaction with Lisa, including when validating results and thus continuously improving potential"
"Lisa uses NLP (natural language processing) algorithms to read texts at infinitely greater speed than human beings and analyses them to identify patterns of meaning, searches for correlations objectively"
"has shown remarkable 'predictive capabilities' that can identify future trends far in advance"
"successfully designed and launched Lisa (Linguistic Intelligence for Supervisory Awareness), a machine learning tool that aids in quickly reading and processing thousands of publications on banking supervision from selected sources such as institutions, think-tanks and consultancy firms."
"the machine allows Intesa Sanpaolo's experts to have a broader view, useful for taking a strategic view for understanding trends and developments."
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
In-house Sample May 12, 2026 Deutsche Bank Deutsche Bank needed to review third-party vendors' documentary evidence against its control framework and regulatory expectations faster without removing expert judgement, because onboarding and periodic reviews… A manual review would typically take around 30 minutes per evidence document. TPRM AI In-house Sample Oct 20, 2025 BNY BNY wanted to scale AI adoption across the company without fragmented point solutions, while maintaining strong governance, risk management, compliance controls, and practical support for workflows such as payments,… 75% more AI solutions in production compared to the prior quarter Contract Review Assistant · Eliza · Eliza agents In-house Sample Oct 16, 2025 Generali Switzerland Claim handlers were managing dozens of cases each day while manually navigating documents, guidelines, and deadlines, which slowed claims processing and left less time for higher-value decision-making. 50 claims of varying complexity KAssis 