Top 10 AI Prompts and Use Cases and in the Financial Services Industry in Italy

Italy's financial services scene is moving fast: a recent

Minsait AI in Italy 2025 report

study finds 63% of large companies have adopted or plan to adopt AI, with a potential productivity uplift of about €115 billion - a signal that banks, insurers and fintechs can turn prompts into measurable value quickly.

At the same time, Banca d'Italia - working with the European Commission and the OECD - has launched a dedicated project to map the opportunities and risks of AI in Italian financial markets, with a final report due in spring 2026, underscoring the need for practical, compliant deployments (Banca d'Italia AI project announcement).

These twin forces - commercial upside and active supervision - make Italy an ideal laboratory for prompts and use cases like fraud detection, dynamic credit scoring and document AI, and a strong reason for finance professionals to build hands-on prompt skills via courses such as Nucamp's Nucamp AI Essentials for Work bootcamp syllabus, a 15‑week practical program that teaches prompt writing and workplace AI applications.

Bootcamp	Length	Cost (early bird)
AI Essentials for Work	15 Weeks	$3,582
Courses: AI at Work: Foundations; Writing AI Prompts; Job Based Practical AI Skills - Register for Nucamp AI Essentials for Work (registration)

Selection of the prompts and use cases followed a triage that favoured official supervision, empirical evidence and hands‑on prompt tactics: priority went to the Banca d'Italia–led EC/OECD project (which combines a nationwide survey of banks, institutional investors and market infrastructures, workshops with regulators and industry, and a final OECD report due in spring 2026) as the primary source of policy and market context (Banca d'Italia artificial intelligence project in the Italian financial sector); technical depth on credit‑scoring practices came from the Bank of Italy's Occasional Paper No.

721, which uses survey data to map AI/ML use and flags issues such as algorithmic bias while noting no clear evidence of greater discrimination versus traditional methods (Bank of Italy Occasional Paper No. 721 - Artificial Intelligence in Credit Scoring (empirical analysis)); and practical prompt design relied on finance-focused playbooks that break complex reporting tasks into micro‑steps, exemplified by DFIN's prompt tips for financial reporting (DFIN practical AI prompts for financial reporting (prompt design playbook)).

The result: prompts and cases that are regulatorily attuned, empirically grounded and immediately testable in Italian bank and insurer settings - imagine swapping a week of manual reconciliation for a validated micro‑prompt that surfaces anomalies in minutes.

Source	Role in methodology
Banca d'Italia AI project	Policy context, nationwide survey, workshops, OECD final report (spring 2026)
Bank of Italy Occasional Paper No. 721	Empirical analysis of AI/ML in credit scoring; risks and survey findings
DFIN knowledge hub	Practical prompt designs and stepwise prompt strategies for financial reporting

HSBC's push into real‑time fraud detection - including the Dynamic Risk Assessment system co‑developed with Google - offers a clear playbook for Italian finance teams that need to cut manual reviews and speed up AML investigations: the bank reports screening well over a billion transactions a month and, after shifting from rules‑only monitoring to ML‑driven models, identified 2–4× more suspicious activity while reducing false alerts by about 60%, bringing time‑to‑detection down to days rather than weeks (HSBC Dynamic Risk Assessment AI fraud detection case study, Google Cloud case study: HSBC AML AI deployment and results).

For Italian banks and supervisors balancing efficiency with explainability, these outcomes underline the value of hybrid deployments (rules + ML), continuous model retraining, and strong governance - practical steps that map onto recent Italian regulatory guidance captured in the Nucamp briefing on IVASS updates (IVASS regulatory updates, April 2025).

The result is tangible: swap a week of manual triage for AI that surfaces networks of suspicious flows in the time it takes to refill a coffee cup.

Metric	Value	Source
Transactions screened monthly	~1.2 billion	Google Cloud case study
False alerts reduction	~60%	Google Cloud case study
Increase in detected suspicious activity	2–4×	Google Cloud case study
Detection time after first alert	~8 days	Google Cloud case study

“The speed itself is mind blowing. You should have seen the faces of some of our guys when they saw the numbers come out in 15 minutes.” - Ajay Yadav, Global Head of Fixed Income for Traded Risk, HSBC

Credit risk assessment is moving from static scorecards to dynamic, explainable models that Italian lenders can use to underwrite more fairly and at scale: Zest AI automated underwriting platform promises 20%+ risk reduction while lifting approvals (25%+ in some cases) and automating up to ~80% of decisions, with proofs‑of‑concept that can run in as little as two weeks - a fast route for banks and fintechs navigating IVASS model governance regulatory updates for Italy 2025.

The practical payoff is clear: better risk segmentation from hundreds of features and SHAP‑style explainability that regulators and credit officers can inspect, meaning more approved applicants without taking on hidden tail risk - effectively turning hours of manual review into instant, auditable decisions that expand access to credit for underserved consumers.

Metric	Value	Source
Population coverage	Assess 98% of adults	Zest AI underwriting
Risk reduction	20%+	Zest AI underwriting
Approval lift	~25% (without added risk)	Zest AI underwriting
Auto‑decision rate	~80%	Zest AI underwriting
Time savings	Up to 60% in lending process	Zest AI underwriting

“Beforehand, it could take six hours to decision a loan, and we've been able to cut that time down exponentially.” - Anderson Langford, Chief Operations Officer

Telecom Italia (TIM) shows how voice agents can move from experiment to everyday efficiency for Italian finance: by modernizing core systems on Google Cloud - migrating some 6,600 internal workloads, shifting ~26% of private‑cloud solutions and using Kubernetes and Citrix virtual desktops - TIM scaled services like the TIM Academy and then put a Google‑powered voice agent into production, boosting contact‑handling efficiency by about 20%.

For banks and insurers in Italy this matters practically: a grounded, omnichannel voice agent can handle routine balance checks, payments queries and identity flows across Italian dialects, cut average handling time, and hand over rich context to human agents for complex cases, freeing staff for high‑value work and improving customer satisfaction.

Read TIM's cloud transformation for infrastructure and training lessons and the Gemini at Work write‑up for the agent‑efficiency example to plan a safe, auditable rollout in regulated settings.

Metric	Value	Source
Contact‑handling efficiency	~20% improvement	Google Cloud Gemini at Work - TIM voice agent case study
Internal workloads migrated	6,600	Telecom Italia (TIM) Google Cloud migration case study
Private cloud migrated (share)	26%	Telecom Italia (TIM) Google Cloud migration case study

“It was crucial for us to find the right partner to drive our cloud transformation. We needed flexibility, scalability, and security, as well as continuous innovation and evolving expertise. Google Cloud fits the bill perfectly.” - Mauro Maccagnani, Head of Digital Enterprise Solutions, TIM

For Italian banks, insurers and fintechs wrestling with mountains of invoices, statements and handwritten forms, Google Cloud's Document AI (built on Vertex AI) offers a fast, pragmatic route to automated transaction capture: enterprise OCR that recognises text in 200+ languages and handwriting in 50 languages, a prebuilt Invoice Parser for common fields, and a Workbench that can be fine‑tuned with as few as 10 documents to reach production‑ready accuracy - ideal for pilots across regional branches and dialects.

Connectors to BigQuery and Vertex Search turn parsed documents into queryable datasets for compliance reporting and analytics, while labs and deployment guides show how to wire Document AI into serverless pipelines (Cloud Run, Pub/Sub) so extracted entities land directly in a data warehouse.

New customers can also trial capabilities with Google Cloud's free credits, making it straightforward to turn a boxed stack of paper into searchable rows in BigQuery and move from manual reconciliation to automated transaction capture.

Learn more with the Google Cloud Document AI overview, the Document AI extraction guide on Google Cloud, or an Invoice Parser walkthrough for Google Document AI.

Feature	Benefit for Italian finance teams	Source
Enterprise OCR (200+ languages; handwriting 50)	Handles Italian, regional scripts and handwritten forms	Google Cloud Document AI overview
Document AI Workbench - fine‑tune with ~10 docs	Rapid pilot & higher accuracy with small labelled sets	Document AI extraction overview on Google Cloud
Prebuilt Invoice Parser + BigQuery integration	Automate invoice & claims processing and feed analytics	Invoice Parser guide for Google Document AI

Predictive cash‑flow forecasting and treasury optimisation in Italy become instantly more achievable with Nilus' AI‑first approach: the platform centralises banks, ERPs and PSPs into a reconciled, always‑on ledger, runs bottom‑up 13‑week forecasts, and surfaces explainable drivers so treasury teams can run what‑if scenarios and act before a shortfall hits.

Nilus integrates with thousands of financial endpoints (the product and best‑practice guides note connections to 20,000+ banks, ERPs and gateways), can be live in days to a few weeks, and is credited with dramatic time‑savings (think replacing a day of manual reconciliation each week and 50+ hours monthly) and near‑real‑time accuracy (Nilus cites ~95% actuals‑vs‑forecast accuracy and fast implementation paths).

For Italian banks, insurers and fintechs juggling multi‑currency ledgers and tight regulatory scrutiny, that means turning reactive spreadsheet work into auditable, scenario‑driven decisions - run a 15% collections shock and see the balance impact in seconds, not meetings.

Learn more from the Nilus cash flow forecasting product overview and the Nilus cash flow forecasting best-practices guide.

Metric	Value	Source
Connected sources	20,000+ banks, ERPs, PSPs	Nilus cash flow forecasting best-practices guide
Implementation time	24 hours – 4 weeks	Nilus cash flow forecasting product overview
Forecast accuracy	~95% actuals vs forecast	Nilus cash flow forecasting product overview
Time saved	50+ hours monthly	Alloy treasury automation case study (50+ hours saved monthly)

“Nilus automated and optimized our treasury planning - outperforming our manual spreadsheet workflows. I use the platform daily to give me the insights into cash positions, cash performance and better forecasting.” - Hai Kim, VP Finance

Speeding month‑end from chaos to confidence is now a practical play for Italian finance teams: automated reconciliation and continuous‑close workflows stitch bank feeds, PSPs and ERPs together, generate audit‑ready journal entries, and surface exceptions for rapid resolution - turning what used to be a week of firefighting into a morning of exceptions‑only reviews (some teams cut 5–6 days from the close in pilots).

Tools that match transactions with AI, centralize task ownership and keep immutable approval logs make audit readiness straightforward for lenders and insurers operating under IVASS and Banca d'Italia scrutiny; see practical automation patterns such as automated transaction matching and real‑time posting in Ledge's month‑end playbook and the implementation guidance found in modern close vendor writeups.

The payoff in Italy is tangible: fewer manual adjustments, clearer sampling for auditors, and time reclaimed for treasury and forecasting instead of pivot‑table triage - imagine handing auditors a single verified ledger instead of a stack of spreadsheets and chasing emails.

Close type	Typical time to close
Small business (manual)	7–10 business days
Mid‑market (partial automation)	4–7 business days
High‑performing (full automation)	1–3 business days

“Successful companies establish clear roles, leverage automation, and treat the close as an ongoing workflow rather than a monthly fire drill.” - David Dolmanet, Director of Finance & Accounting Services

Intesa Sanpaolo's approach to regulatory compliance, AML monitoring and explainability pairs industrial‑scale data platforms with targeted anti‑financial‑crime governance - building a “Democratic Data Lab” on Google Cloud that speeds release of risk solutions by 20–30%, cuts stress‑test reporting time by up to 80% and gives the board real‑time visibility via Looker dashboards (Intesa Sanpaolo Google Cloud case study).

That cloud‑first lab and the bank's AI research arm (AI Lab) feed explainable pipelines and automated extraction with Gemini so analysts spend less time wrangling data and more on oversight, while group policies and training underpin AML/CFT controls (Intesa Sanpaolo AML/CFT regulatory policies).

Yet operational reality is sobering: an alleged internal access incident affecting ~3,500 accounts in 2022–24 highlights why granular access monitoring, anomaly detection and transparent model explainability are non‑negotiable for Italian finance - practical safeguards that turn black‑box alerts into auditable evidence for supervisors and investigators (Daily Security Review: Intesa Sanpaolo data breach incident summary).

Metric	Value
Faster release of risk solutions	20–30%
Time to complete EU stress tests	Reduced by up to 80%
Data extract/load time saved (Gemini 1.5)	~30%
Customers / Branches (at a glance)	13.7 million / 3,300+
People trained on AML/anti‑corruption (2024)	88,109

“We have defined metrics on all sorts of risks at the bank and we set early warning alerts to help us manage those risks effectively… the risk committee is constantly aware of what is happening in real time.” - Domenico Fileppo

Operationalising models at scale is where MLOps stops being theory and starts saving time and risk: Vertex AI Pipelines stitches data ingestion, training, batch inference and automated model evaluation into repeatable runs so teams can spot drift, compare versions and gate deployments without manual firefights.

For Italian insurers and banks - and organisations like Generali Italia considering cloud MLOps - the built‑in evaluation tooling surfaces the exact metrics supervisors care about (precision, recall, AuPRC, confusion matrices for classification; MAE/RMSE/MAPE for regression and forecasting) while computation‑based evaluation pipelines let teams evaluate LLM outputs against prompt/ground‑truth pairs (BLEU, ROUGE‑L, Exact Match) and require only a small, well‑curated JSONL test set to start.

Make model checks automatic: trigger a pipeline when new ledger data lands, run batch predictions, and publish versioned evaluation artifacts to Cloud Storage and the Model Registry so auditors see the full lineage in seconds, not weeks.

Learn the evaluation basics in the Vertex AI model evaluation overview - Google Cloud and the Vertex AI computation-based evaluation pipeline for LLMs - Google Cloud.

Task	Key metrics	Source
Classification (tabular)	Precision, Recall, AuPRC, AuROC, F1, Confusion matrix	Vertex AI model evaluation overview - Google Cloud
Regression / Forecasting	MAE, RMSE, RMSLE, r², MAPE, WAPE	Vertex AI model evaluation overview - Google Cloud
LLM / Text generation	BLEU, ROUGE‑L, Exact Match (task dependent)	Vertex AI computation-based evaluation pipeline for LLMs - Google Cloud

Credem shows how personalization at scale looks in Italy: Credem's virtual agent Emily blends LLMs and advanced NLP to answer customers 24/7 on the website and WhatsApp, perform on‑the‑spot loan calculations, geolocate ATMs and propose the right current account or product based on context and follow‑ups (Credem Emily virtual agent project by Andemili); that front‑line intelligence is paired with deep automation and process mining that slashes manual time and frees teams to craft tailored offers - IBM's work with Credem cut employee loan‑processing time by about 70% and supported 91 automations saving roughly €1.4M (IBM process mining and automation case study with Credito Emiliano).

On the channel side, a Temenos‑powered mobile app and low‑code approach let Credem iterate fast and deliver highly personalized in‑app journeys, driving sustained digital engagement (Temenos Credito Emiliano SPA success story).

The upshot for Italian banks: combine conversational AI, backend automation and rapid app releases to nudge the right product to the right customer - local touches matter, too (yes, Credem famously accepts Parmigiano‑Reggiano as collateral), and that cultural resonance can turn a marketing nudge into loyalty.

Metric	Value	Source
Active app users	~550,000	Temenos Credito Emiliano SPA success story
Digital transactions share	95%	Temenos Credito Emiliano SPA success story
Employee time on loan processing	Reduced ~70%	IBM process mining and automation case study with Credito Emiliano
Automations implemented	91 (savings ≈ €1.4M)	IBM process mining and automation case study with Credito Emiliano

“With Temenos, we have been able to quickly develop a mobile banking experience to compete with the very best in Italy.” - Fabio Caliceti, Head of Digital Channels, Credito Emiliano

Internal virtual assistants and no‑code chatbots are an immediate productivity lever for Italian finance teams - think instant answers to reconciliations, policy queries and onboarding checklists without waiting for IT. Platforms like Denser.ai no-code chatbot platform let non‑technical finance staff build assistants that pull answers from internal docs, highlight the source of every reply for auditability, understand tables and charts, and connect into Slack or other workflows so approvals and handoffs happen inside the tools teams already use; that source‑transparent behaviour is especially useful when auditors or regulators need traceable evidence.

For internal use cases - from helpdesk FAQs and access provisioning to expense-query resolution - research shows chatbots cut repetitive work and speed decision‑making, turning a two‑hour policy hunt into an answer in the time it takes to pour an espresso.

For teams that need inspiration on proven internal workflows and ROI, see real‑world patterns collected in industry writeups on internal AI chatbots and no‑code deployment best practices (Master of Code internal AI chatbots best practices).

Plan	Price (month)	Included bots / queries
Free	$0	1 DenserBot / 20 queries
Starter	$19	2 DenserBots / 1,500 queries
Standard	$89	4 DenserBots / 7,500 queries
Business	$799	8 DenserBots / 15,000 queries

Getting started with AI in Italy's financial sector is less about chasing the latest model and more about pairing modest, measurable pilots with clear governance and skills-building: follow the Bank of Italy's innovation signals (the Milano Hub call for proposals that targets AI for banking, payments and financial inclusion) and heed the ECB's reminder that AI brings both efficiency gains and systemic risks unless data, model development and deployment are managed carefully (Bank of Italy Milano Hub initiatives, ECB Financial Stability Review on AI).

Start with high‑value, low‑risk use cases - document AI pilots, reconciliation automation or constrained credit‑scoring proofs‑of‑concept - measure outcomes, document explainability and scale only once governance and auditors are satisfied.

Parallel to pilots, invest in human capital so teams can write effective prompts, evaluate model outputs and spot bias: practical courses such as Nucamp's 15‑week AI Essentials for Work teach prompt design and workplace AI skills that speed adoption while keeping compliance front of mind (Nucamp AI Essentials for Work syllabus), turning months of manual effort into minutes of auditable insight.

Bootcamp	Length	Early bird cost
AI Essentials for Work - practical prompt & workplace AI skills	15 Weeks	$3,582 - Register for Nucamp AI Essentials for Work