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

Cleveland's financial-services cluster is ripe for AI because regionally headquartered banks, fintechs and growing tech infrastructure already converge here: KeyBank - headquartered in Cleveland with approximately $189 billion in assets - anchors a 15‑state regional footprint and is deploying machine learning in commercial products, most recently the KeyTotal AR predictive accounts‑receivable platform that automates cash application and claims up to 50% AR cost savings (KeyTotal AR predictive accounts-receivable platform press release); Ohio's competitive IT talent pool and major cloud investments make production-ready pilots more affordable than in New York or SF, and practical workforce programs like Nucamp's Nucamp AI Essentials for Work syllabus teach nontechnical staff how to write prompts and embed AI across operations - so Cleveland firms can quickly move from pilots to measurable savings, not just experiments (KeyCorp company overview and corporate information).

Metric	Value
Headquarters	Cleveland, OH
Assets (3/31/25)	~$189 billion
Full-time employees	~17,000
Branches	~1,000

"In today's uncertain environment, our clients are looking to cash flow optimization to help them grow their business. Working with innovators like Versapay enables us to optimize clients' cash flow and operations from beginning to end." - Lindsay Weinstein, Head of KeyBank's Merchant Services

Methodology: selection emphasized Cleveland relevance, measurable pilot outcomes, and workforce impact - prompts and use cases were chosen when a local proof point or operational metric showed clear business value (for example, Fuel50's Grow at Key reported a 72% user return rate and 9,858 skills assessed, signaling sustained employee engagement) and when conversational pilots demonstrated containment or efficiency gains (KeyBank's MyKey logged ~3,000 daily sessions in its first month, 250,000 interactions and an 84% containment rate).

Priority criteria: headquarters or regional deployment in Ohio, documented pilot metrics, human-in-the-loop design for reskilling and ethical oversight, cross-functional teams for productization, and the “start small, scale” operational approach KeyBank used in MyKey.

Sources include detailed reporting on KeyBank's AI and talent programs and KeyBank's own MyKey documentation, used to map prompts to measurable outcomes and Cleveland's hiring and reskilling needs (Emerj analysis of Artificial Intelligence at KeyBank, KeyBank MyKey virtual assistant official page).

Metric	Value
Fuel50 (Grow at Key) user return rate	72%
Skills assessed under Future Ready	9,858
MyKey daily sessions (first month)	~3,000
MyKey interactions logged	250,000
MyKey containment rate	84%

"We probably hired less because we have less call volume."

KeyBank's Grow at Key uses Fuel50's AI-powered internal talent marketplace to make skills visible across the Cleveland organization, matching employees to gigs, courses and roles so HR can redeploy talent rather than default to external hires; the program's concrete outcomes - 72% user return rate, 9,858 skills assessed and 2,774 upskilling/reskilling actions - show how an AI prompt that surfaces “best-fit” internal opportunities turns reskilling from a one-off into measurable workforce capacity-building for Ohio teams (Fuel50 Grow at Key case study) and mirrors the broader strategy analysts observed in KeyBank's AI playbook (Emerj analysis of AI at KeyBank).

For Cleveland financial-services leaders, the practical takeaway is simple: embed an AI marketplace prompt that recommends development actions and internal matches, and talent pipelines become a predictable lever for retention and operational agility.

Metric	Value
User return rate (Grow at Key)	72%
Skills assessed (Future Ready)	9,858
Upskilling/reskilling actions set	2,774
Aspiring Leaders participation increase	100%
Training participation increase	60%

"Our original intent was always about increasing internal mobility, driving employee engagement, increasing retention, and to really help position KeyBank to continue to deliver for our clients." - Carole Torres, SVP & Chief Learning Officer

KeyBank's MyKey conversational‑agent support prompt acts as a real‑time knowledge companion for Cleveland customer‑service teams: built on Google AI and embedded in the mobile app, MyKey logged roughly 3,000 daily sessions in its first month, 250,000 interactions and an 84% containment rate - metrics that reduced call volume and produced measurable productivity gains for contact centers.

The practical prompt pattern to replicate locally: surface account context, return concise recommended next steps (scripts or transfers), and auto‑populate escalation notes so agents resolve more complex cases faster while routing unresolved issues to humans to preserve compliance and trust.

For implementation detail and outcomes, see KeyBank's MyKey documentation and an industry analysis of KeyBank's AI rollout (KeyBank MyKey virtual assistant documentation, Emerj analysis of KeyBank AI rollout).

Metric	Value
Daily sessions (first month)	~3,000
Interactions logged	250,000
Containment rate	84%

"We probably hired less because we have less call volume."

Bank of America's Erica virtual assistant brings practical self‑service to Ohio customers through the Mobile Banking app - answering account questions, searching transactions, notifying users about FICO® score changes, and offering live chat or specialist handoff when needed; critically for Cleveland banking moments, Erica can temporarily lock or unlock a misplaced debit card, help replace a lost or stolen card, and track refunds so customers don't need an in‑person visit (Bank of America Erica virtual assistant features).

Because Erica is available only inside the authenticated Mobile App and uses natural language processing with scripted responses, interactions are logged for quality and routed to humans when the assistant can't resolve an issue - helpful when suspicious activity arises; follow the bank's fraud reporting and recovery steps if a compromise is suspected (Bank of America replace debit card and enable digital card, Bank of America report suspicious activity and fraud recovery).

The practical payoff: Cleveland users can lock a card and switch to a digital token instantly, reducing fraud exposure and avoiding a branch trip.

Erica Capability	Notes
Manage cards	Lock/unlock, replace lost/stolen cards, digital card access
Account insights	Balances, spending categories, recurring charges
Transaction tools	Search transactions, track refunds, Zelle® support
Security & routing	Authenticated in Mobile App; recorded chats; live chat/escalation

Design a fraud‑detection triage prompt that mirrors HSBC's ML approach: score incoming transactions and account activity in real time, surface network links and exception narratives, then prioritize alerts by risk so Cleveland compliance teams investigate fewer, higher‑quality cases instead of chasing noise - HSBC's AML AI cut false positives ~60% while flagging 2–4× more suspicious activity and sped reviews to days not weeks (screens ~1.2–1.35B transactions monthly) (HSBC AML AI case study on Google Cloud).

Pair the triage prompt with automated SAR drafting and human‑in‑the‑loop labeling to continuously retrain models and lower local compliance costs, a pattern other vendors report as cutting alerts and operational burden by up to 60–70% (smart AML software cost and efficiency findings); the practical payoff for Cleveland banks: fewer false alarms, faster investigations, and more analyst time for complex, high‑impact cases.

Metric	Reported result
False positive reduction (HSBC)	~60%
Transactions screened monthly	~1.2–1.35 billion
Suspicious activity detection uplift	2–4×
Review time after ML	Reduced to ≈8 days

"[Anti-money laundering checks] is a thing that the whole industry has thrown a lot of bodies at because that was the way it was being done. However, AI technology can help with compliance because it has the ability to do things human beings are not typically good at like high frequency high volume data problems." - Andy Maguire, HSBC Chief Operating Officer

Design a credit‑decision support prompt for Cleveland banks that combines FCRA‑compliant alternative data (rent, utilities, cellphone payments) with explainable ML outputs so underwriters see a compact applicant summary, a Shapley‑based “principal reasons” explanation for adverse actions, and automated monitoring alerts for input drift, reason‑code instability, and latency - a pattern Zest AI recommends to expand fair access while satisfying examiners (AI lending data documentation and monitoring best practices).

Include a human‑in‑the‑loop review gate for thin‑file Ohio applicants and a prompt that suggests legally defensible alternative models when disparate impact is detected; Zest's client results (e.g., 10% card, 15% auto, 51% personal‑loan approval lifts with no increase in defaults) show the “so what”: local credit unions and community banks can responsibly increase approvals for thin‑file Cleveland residents without raising loss rates (Zest AI federal guidance commentary and approval lift results).

Practical rollout advice: start with a decision‑support prompt that outputs score, top three drivers, compliance reason codes, and a retraining trigger so pilots move quickly into auditable production.

Metric	Reported result
Credit card approval lift	10%
Auto loan approval lift	15%
Personal loan approval lift	51%

“Bank management should be aware of the potential fair lending risk with the use of AI or alternative data... It is important to understand and monitor underwriting and pricing models to identify potential disparate impact and other fair lending issues. New technology... such as machine learning, may add complexity while limiting transparency. Bank management should be able to explain and defend underwriting and modeling decisions.” - OCC

Design an AML/KYC monitoring prompt for Cleveland banks that ingests live transaction streams, KYC profiles and device signals, then runs a hybrid rule+ML pipeline to score and prioritize alerts - surface network links for fast‑moving layering or mule networks, attach a risk rationale and an auto‑populated SAR draft for FinCEN (U.S. SAR windows ≤30 days), and route high‑risk cases to enhanced‑due‑diligence queues for human review; this pattern combines Sumsub's real‑time vs.

batch guidance with continuous screening and case management to cut manual triage and improve signal quality (Sumsub AML transaction monitoring guide, SanctionScanner transaction monitoring fundamentals and U.S. SAR timing).

Include ongoing re‑screening, fuzzy matching for sanctions/PEP hits, and one‑click enrichment from customer‑screening APIs so local compliance teams spend 25% fewer hours on low‑value alerts and more on complex investigations - a tangible way Cleveland banks and credit unions reduce regulatory risk while scaling digital services (SEON customer screening and monitoring).

Feature	Why it matters
Real‑time monitoring	Prevents live laundering schemes and sanctions breaches
Hybrid rules + ML	Balances explainability with adaptive detection
Network analysis	Reveals mule rings and hidden beneficiary links
Automated SAR drafting	Speeds filing within U.S. 30‑day deadlines
Ongoing screening	Reduces false positives and keeps KYC current

“It is a robust analytics and risk management system that offers customizable options for setting relevance thresholds and reducing false positives.” - Monika Zaja, Fraud Manager, YSI

For Cleveland banks and credit unions facing dense regulatory filings, long vendor contracts, and voluminous discovery, a document‑summarization & compliance prompt built on finance‑and‑law tuned models can cut review time while preserving audit trails: Bloomberg Law's AI Assistant and Brief Analyzer accelerate research and generate summarized legal language with citations, and BloombergGPT - Bloomberg's finance‑focused LLM - adds domain grounding for market and regulatory content (Bloomberg Law AI Assistant & Brief Analyzer for legal research and brief analysis, BloombergGPT finance LLM: training and finance industry use cases).

Independent benchmarks show AI tools outperform lawyers on summarization and deliver dramatically faster results, so a Cleveland prompt should: accept multi‑document uploads (contracts, SEC filings, SAR narratives), return concise summaries with cited passages and compliance reason codes, and flag items for human review to meet ethical and exam‑readiness standards (VLAIR legal AI benchmark for summarization and comparative results).

Model / Tool	Relevant capability
Bloomberg Law AI Assistant	Fast legal research, summarization, citation generation
BloombergGPT	Finance‑tuned LLM for document interpretation and sentiment

“Six times faster than the lawyers at the lowest end, and 80 times faster at the highest end.”

A portfolio‑construction prompt for Cleveland asset managers and RIAs can automate the hard tradeoffs between thematic opportunity and portfolio stability: prompt inputs (target client risk, horizon, taxable status) should return a recommended thematic tilt, suggested funding source (sector, style, or new cash), expected unique‑risk contribution, and a batch of Aladdin‑style scenario runs to stress test outcomes under Fed, inflation, or geopolitics shocks.

Use BlackRock's thematic framework to size themes - remember: thematic exposures show higher unique risk (BlackRock notes thematic unique risk ≈30% versus sector ≈24% and style 16–19%) but can materially drive returns (AI‑related stocks contributed nearly half of S&P 500 returns in 2024) - and pair that with Aladdin's whole‑portfolio scenario tooling to produce explainable P&L paths and attribution for client conversations.

The practical payoff for Cleveland firms: one prompt that outputs a funded trade plan, scenario P&Ls, and risk attributions turns thematic ideas into auditable, client‑ready allocations tied to measurable stress outcomes (BlackRock thematic investing guide, BlackRock Aladdin market-driven scenario tools).

Metric	Reported value
Thematic unique risk (example)	≈30%
Sector unique risk (example)	≈24%
Style unique risk (value/growth)	≈16% / ≈19%
Aladdin risk factors	3,000+

“investors should focus more on themes and less on broad asset classes as mega forces reshape whole economies.”

A Customer Personalization prompt for Cleveland banks should combine transaction signals, app behavior and life‑event indicators into a real‑time next‑best‑offer decision: surface a contextually relevant product or message in the mobile session (e.g., a mortgage checklist after repeated rent payments or a travel card offer after a surge in airfare purchases) so offers land when intent is highest; predictive analytics and Agentic AI frameworks explain how to score and time those moments (predictive insights and Agentic AI for banks), while industry benchmarks show customers expect and value contextual offers and that recommendation engines materially drive revenue (Blend research on personalization in banking, Dynamic Yield personalization engine results).

The so‑what: real‑time, propensity‑based prompts turn routine app activity into measurable outcomes - firms that tie offers to live signals typically see much higher conversion and engagement than broadcast campaigns, making a focused personalization prompt one of the fastest paths from pilot to ROI for Cleveland institutions.

Metric	Reported value
Customers who value contextual offers	≈75% (Blend)
Revenue driven by recommendations	≈25% (Dynamic Yield)
Real‑time campaign conversions (example)	up to 66% (Techcombank / Adobe)

Cleveland's recent cybersecurity wake‑up call - two cyberattacks in under a year and an Ohio Auditor finding that some departments (including the Municipal Court and Department of Public Utilities) lacked consistent MFA and regular security review - makes an incident‑response prompt for login anomalies urgent: combine device‑fingerprinting signals (hardware/software telemetry, canvas/audio codecs, OS and network attributes) with AI‑driven behavioral cues (ML that can monitor thousands of login signals) to flag impossible‑travel, velocity spikes, and session‑hijacking attempts, then automatically trigger adaptive MFA, session revocation, and a human review queue; device fingerprints also support “trusted device” decisions that reduce customer friction when safe.

Practical playbook items drawn from industry guidance: deploy device fingerprinting as one layer of fraud detection (see Device fingerprinting for fraud detection - Transmit Security), mandate and streamline MFA adoption per FINRA best practices for ATO prevention (see FINRA guidance on MFA and account takeover prevention), and bake the IR prompt into playbooks so analysts get prioritized, high‑quality alerts instead of noise (see Ohio Auditor Cleveland cybersecurity assessment).

Local finding	Mitigation in prompt
2 cyberattacks in one year	Automated anomaly detection + rapid containment playbook
Inconsistent MFA; courts lacked MFA	Adaptive MFA enforcement and trusted‑device workflows
Need for richer signals	Device fingerprint telemetry + behavioral ML for higher signal quality

"We've significantly boosted our security, including full Multi-Factor Authentication (MFA) implementation for all court employees."

For Cleveland financial-services teams ready to move from experimentation to impact, the practical path is clear: pick one high-value prompt (fraud triage, AML/KYC, or a customer‑facing assistant), run a measurable pilot that ties outcomes to reduced analyst hours or containment rates, and simultaneously upskill staff so the organization can operationalize models rather than chase proofs‑of‑concept; practical how‑tos for turning metrics into decisions are usefully summarized in the Smartsheet Quick-Start Guide to Data-Driven Decision Making (Smartsheet Quick-Start Guide to Data-Driven Decision Making).

Anchor pilots with systems‑level modeling and lifecycle thinking to avoid stove‑piping risk and ensure auditability (see the PPI SyEN systems engineering newsjournal, issue 94 (PPI SyEN systems engineering newsjournal, issue 94)), and pair that with practical workforce training - Nucamp's 15‑week AI Essentials for Work program teaches prompt writing and real workplace application so nontechnical staff can run and validate pilots (Nucamp AI Essentials for Work syllabus).

The so‑what: combine short, measurable pilots with a concrete 15‑week upskilling cadence and Cleveland teams can convert AI from a cost center into repeatable operational savings and faster, auditable decisions.

Program	Length	Early‑bird Cost	Registration
AI Essentials for Work	15 Weeks	$3,582	Register for Nucamp AI Essentials for Work (15 Weeks)