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

Colombia's financial sector is racing to turn AI from an efficiency play into a tool for inclusion: Bancolombia Ventures‑backed Quipu uses alternative data and AI to score microbusinesses - reviving an idea as clever as the Incan quipu - to lend tiny amounts ($100–$2,000) that have added up to about $3.5M and reach firms in a market where roughly 95% of businesses are microenterprises that employ 65% of the workforce (Quipu AI credit-scoring for microbusiness lending).

Across the region, generative AI is reshaping fraud detection, personalization and real‑time automation for banks (Generative AI transforming fraud detection and personalization in Latin American banking), while Colombia's regulatory sandbox and Open Finance push create room to experiment safely (Colombia regulatory sandbox and Open Finance policy enabling fintech experiments).

Practical workplace AI skills - like prompt design and safe deployment - are essential for Colombian teams to scale these wins without amplifying bias or risk.

Bootcamp	Length	Cost (early / after)	Registration
AI Essentials for Work	15 Weeks	$3,582 / $3,942	Register for AI Essentials for Work bootcamp

“Loan sharks were these businesses' only solution. We're an alternative to that.”

To assemble the “Top 10” use cases for Colombia's financial sector, sources were scanned for regionally relevant, production-ready examples and for frameworks that make adoption repeatable: Google Cloud's expanded compendium - now 601 real-world gen‑AI use cases - provided the industry taxonomy (Customer, Employee, Data, Security, Code, Creative) and many concrete finance examples (Google Cloud 601 real-world generative AI use cases compendium), while the product-level guidance on how to frame problems, build retrieval-augmented pipelines, and follow Google's responsible‑AI guardrails helped shape selection and prompts (Google Cloud generative AI examples and responsible-AI guidance).

Selection criteria prioritized: clear applicability to Colombia (e.g., Colombian Security Council chatbot, Habi's document automation), measurable impact (Banco Covalto's >90% faster credit responses as a regional precedent), regulatory and data‑governance fit, and promptability for non‑technical teams.

Example prompts and templates came from the provider playbooks (searchable prompt examples and “What problem are you trying to solve?” starter questions), plus Nucamp's local briefing on credit‑scoring and fraud use cases to ensure recommendations map to Colombian realities (AI credit-scoring and document automation in Colombia financial services), so each use case here is chosen for practical payoff and safe, testable deployment.

Customer-facing multilingual virtual assistants on WhatsApp are fast becoming the frontline for Colombian banks and fintechs because they meet customers where they already live - on a channel with 80–90% regional penetration - and they cut friction for low‑digital‑literacy users: Tpaga and other players let customers open accounts and pay bills inside chat, while inclusive lenders built around informal economies use WhatsApp to collect applications and media evidence that traditional forms miss, speeding decisions to “within a day or three” (a game‑changer for microbusinesses).

Beyond convenience, conversational banking boosts conversion and lowers costs - product campaigns via WhatsApp can lift conversion by up to 30% and conversational channels can reduce call‑centre volume by around 40% - and the same chat thread becomes a traceable record for onboarding, payments and dispute resolution.

For Colombian institutions exploring a multilingual assistant (Spanish, regional dialects and even voice notes), the practical playbook is clear: use WhatsApp flows as the UX layer, pair NLP with identity checks and human escalation, and partner with certified providers to stay compliant while harvesting high‑signal behavioural data.

See why WhatsApp is now talked about as the region's fintech “operating system” and how conversational banking delivers measurable ROI in practice (WhatsApp as the fintech operating system - Emerging Fintech analysis, Conversational banking use cases in Latin America - Latinia resources) and how informal‑economy lenders route applications via chat in Colombia (Quipu WhatsApp applications case study - How to Lend Money to Strangers podcast).

“Everybody knows how to use WhatsApp.”

Automated underwriting is emerging as a practical lever for Colombian lenders to expand small‑loan access while cutting turnaround from days to minutes: regional precedents such as Banco Covalto's >90% faster responses show what's possible when API‑driven data collection replaces paper trails and human bottlenecks.

Modern systems automate four key steps - rapid data collection via open‑banking APIs, machine‑led analysis and scoring, calibrated risk assessment, and rule‑based decisioning - so smaller loans can be approved the same day or even in minutes rather than weeks (Automated underwriting for SMBs - Canopy).

Practical builds used by innovators combine real‑time transaction feeds, LLM‑assisted data enrichment, and modular “decision engines” (data, scoring, rules, fraud, product) to keep risky edge cases routed to human review while automating the routine 85% of approvals - Defacto reports outcomes like funds in six minutes versus an industry average of three weeks (Real-time underwriting and decision engines - Defacto).

For Colombian banks and fintechs, the playbook is straightforward: plug into reliable data sources, instrument transparent decision rules for auditability, and pilot with clear human checkpoints so faster credit becomes inclusive and defensible (AI Essentials for Work bootcamp syllabus - Nucamp).

Fraud detection and AML automation for Colombian banks is moving toward real‑time, risk‑based pipelines that blend transaction monitoring, behavioural signals and jurisdictional scoring so alerts arrive as actionable leads instead of noise; modern systems surface unusual velocity, cross‑border flows and device anomalies instantly and can route the riskiest 5–10% of cases to investigators while automating low‑risk triage to protect customer experience.

Implementers should expect integration and cost hurdles - legacy core systems and high false‑positive rates remain the main barriers - but firms that combine configurable, no‑code rules with machine‑learning enrichment and country risk inputs can both lower alert volumes and meet regulator expectations.

Practical steps for Colombia: feed country‑risk scores and FATF updates into customer risk ratings, tune thresholds for medium‑risk jurisdictions like Colombia (SanctionScanner's composite score 5.40), and layer sanctions/PEP screening with behavioural fraud signals to reduce manual work and speed SARs to the UIAF when needed (SanctionScanner country risk scoring guide for AML and compliance teams, Alessa's best practices for real‑time AML monitoring, SEON overview of transaction monitoring capabilities).

Country	Composite Score	Risk Level
Colombia	5.40	Medium

“Using AI-powered rules-based transaction monitoring tailored to an institution's risk profile significantly reduces false positive AML alerts, improving operational efficiency.”

Document ingestion, contract analysis and KYC automation are the plumbing that turns messy onboarding into auditable, regulatory‑safe workflows in Colombia: well‑designed pipelines capture PDFs and WhatsApp media, extract parties and clauses, apply tiered e‑signature logic, and produce an immutable audit trail that stands up to both the National Archives' retention rules and courtroom scrutiny.

In Colombia that means mapping every workflow to Law 527/Decree 2364's distinction between Simple Electronic Signatures and government‑accredited Digital Signatures, following the General Archives Law and Decree 1080 for preservation, and building contract‑summary outputs that satisfy transparency obligations under Law 1712 and the new fiduciary disclosure expectations in the draft trust regulation.

Practically, vendors and in‑house teams should combine robust OCR, clause‑level NLP and secure key‑management so high‑value contracts route to Digital Signatures while lower‑risk forms use SES; the result is faster KYC, fewer manual reviews and a searchable, court‑ready record instead of a room full of paper.

See the legal framework for e‑signatures in Colombia and the country's document‑management rules for implementation guidance.

Regulation	Core requirement
Colombia e-signature law 527 and Decree 2364	Defines Simple vs Digital Signatures; Digital Signatures require accredited Trust Service Providers
Colombia document management regulations: Law 594 and Decree 1080	General Archives Law and preservation/EDMS rules for electronic documents
2025 draft fiduciary decree - fiduciary regulation proposal for Colombia	Heightened disclosure, reporting and contractual clarity for fiduciary arrangements

Regulatory reporting teams and in‑house counsel in Colombia can turn AI from a compliance headache into a multiplier - if it's wired to local rules: the Government's July 2025 AI Bill creates a risk‑based regime (critical, high, transparency‑obligated, minimal) and names the Ministry of Science as the national authority while keeping the Superintendencia Financiera (SFC) central to financial oversight (Baker McKenzie analysis of Colombia AI Bill (July 2025)); that matters because the draft authorises steep sanctions (fines up to 3,000 monthly minimum wages, suspensions of up to 24 months or even closure) and requires traceable documentation, transparency and human oversight.

Practical pilots should pair legal‑AI search and memo drafting with operational tools that produce auditable inputs for regulatory reports - think reusable regulatory data dictionaries and automated mappings so a single data model feeds SFC returns, risk assessments and board memos with the same definitions (Nasdaq article on enhancing Colombian regulatory reports and data dictionaries).

Cloud and provider compliance mappings (e.g., SFC mappings from major cloud vendors) help teams understand what can be outsourced and what must remain under tight controls, so legal research, compliance memos and regulatory filings become repeatable, defensible workflows rather than last‑minute scramble (Google Cloud compliance mapping for the Financial Superintendence of Colombia).

Regulatory Element	What it requires
AI risk classification	Classify systems as critical, high, transparency‑obligated, or minimal
Documentation & oversight	Maintain impact assessments, documentation, human oversight and transparency
Enforcement	Fines up to 3,000 minimum monthly wages; suspension/closure up to 24 months

GFT's push into Colombia is reshaping how local banks modernize cores and boost developer productivity: the 2024 Sophos Solutions acquisition gives GFT deep Colombian talent and a fast lane into core‑banking projects, and the company now cites a 50–90% productivity uplift from its GFT AI Impact tooling - a practical lever for teams migrating legacy systems to cloud‑native, API‑first architectures that let product and engineering squads iterate faster and reduce time‑to‑market (GFT 2024 press release on growth and AI Impact tooling).

The buy of Sophos also expanded GFT's LatAm footprint and delivery capacity, helping win a major core‑modernization deal with Colombia's largest bank - proof that local capability plus global patterns (Vault‑style cloud cores, microservices and partnership playbooks) can turn risky, multi‑year migrations into staged, testable sprints that preserve regulatory audit trails and accelerate feature rollout (FinTech Magazine coverage: GFT acquires Sophos Solutions in Colombia).

For Colombian CTOs, the takeaway is concrete: modern cores aren't theoretical - they're a pathway to faster product velocity, clearer auditability and developer workflows that finally scale across branches and fintech partners.

Metric	Value / Note
Group revenue (9M 2024)	EUR 645.50 million
GFT AI Impact productivity	50–90% boost
Sophos acquisition	Effective 1 Feb 2024 - expanded LatAm talent & clients
New deal	Major core banking modernisation at Colombia's largest bank
Sales traction	10 new clients, >170 licenses, >20 POCs (last quarter)

“We have just won a major new deal in core banking modernisation at Colombia's largest bank. We have also achieved some major milestones with our AI solutions...”

United Wholesale Mortgage's automation playbook offers a clear blueprint for Colombian lenders wrestling with paperwork-heavy loan servicing, mortgages and claims: intelligent document processing (UiPath Document Understanding) slashed invoice review from three minutes to roughly 30 seconds and removed 50,000 items from queues, while a Google Cloud partnership doubled underwriter throughput - real-world proof that automation can free staff to focus on exceptions and customer care rather than triage (UiPath Document Understanding case study, UWM–Google Cloud AI partnership).

Pairing IDP with workflow orchestration (the microservices/Orkes pattern that cut service delivery from months to minutes) makes staged pilots safe and auditable - so Colombian servicers can target hours saved per file, not just shiny pilots, and scale without losing regulatory traceability (Orkes Conductor at UWM).

One vivid test: what used to be a three-minute invoice chore becomes a 30‑second verification, turning back‑office drag into a competitive servicing advantage.

Metric	Outcome
Invoice processing speed	3 minutes → 30 seconds (≈84% faster)
Automation at scale	90% for mortgage loan invoices
Work items removed	50,000 fewer queue items
Underwriter capacity	Increased from 6 → 14 loans/day (Google Cloud partnership)

“We built an app with UiPath to extract that data and show it to the team member side-by-side with the invoice, so they could make quick decisions.” - Dustin Spivey, Senior Team Leader of Enterprise Technology

SURA Investments is sharpening personalized advisory and wealth management for Colombian clients by pairing broader product reach with a clear-eyed view of technology and governance: its 2025 outlook names the “rise of Artificial Intelligence (AI)” as a structural market force shaping investment strategy, which signals that algorithm‑augmented personalization may figure into future client offerings (SURA Investments' 2025 outlook); at the same time SURA has expanded on‑the‑ground capacity - completing acquisitions that deepen local fiduciary capabilities - and opened offshore channels so AFP Protección clients can access international funds via SURA's Miami RIA, widening Colombians' access to diversified strategies without leaving the domestic system (AFP Protección partnership with SURA's Miami RIA).

Confidence in these moves is reinforced by strong governance signals - an independent Kroll forensic review found no accounting fraud at Grupo SURA - so wealth teams can pilot personalized, data‑driven advice with a governance backbone that pension and retail clients expect (Kroll forensic report on Grupo SURA); the practical takeaway for Colombian institutions is simple: combine expanded product access, AI‑aware strategy and proven governance to deliver advice that scales while keeping client trust intact.

“We are very pleased to strengthen our investment offering and be able to increasingly address the needs of our clients in a more comprehensive manner.” - Juan David Correa, president of Protección

Back-office teams in Colombia can slash busywork and surface decision-ready intelligence by using Microsoft 365 Copilot to condense long documents, meeting transcripts and file collections into crisp briefs and action lists; Copilot in Word can generate automatic summaries at the top of a document or, on demand, answer follow-up prompts to pull key numbers and next steps (Microsoft Copilot in Word automatic summary feature), while Copilot Tuning lets institutions enforce an organisational voice and run summarization inside a tenant‑isolated environment so sensitive regulatory and client data stay under control (Microsoft Copilot Tuning tenant-isolated summarization for enterprises).

Practical wins are immediate: meeting recaps and chat digests can be produced several times faster, Teams transcripts feed clear action items, and supervised fine‑tuning keeps summaries audit‑ready for SFC reporting or board memos - turning information overload into concise, traceable inputs for compliance and frontline decisions.

“The ability to extract actionable insights from large business documents instantly is revolutionizing corporate workflows.”

For Colombian banks and risk teams, modern data analytics and scenario simulation must marry rigorous stress‑testing with the practical tools that make experiments repeatable: Google's Vertex AI offers built‑in A/B experiment monitoring (traffic‑split checks, per‑search and per‑visit business metrics and confidence intervals) so teams can validate changes to scoring, segmentation or model retrieval pipelines without blind rollouts (Vertex AI Search A/B experiment monitoring documentation).

At the same time, AI‑specific stress frameworks demand forward‑looking scenario generation, synthetic data and exhaustive backtesting so models don't “fail fast” under shocks - Ankura's work shows how credit models that looked stable before COVID suddenly misclassified large cohorts when stress conditions arrived, so simulation must test fairness, calibration and feature fragility, not just accuracy (Ankura stress‑testing AI models for model risk management).

Practical pilots echo Strise's playbook for production: use Vertex AI's LLMs to enrich signals, automate continuous KYC/AML triggers, and integrate explainability and IAM controls so alerts become actionable rather than noise (Strise uses Vertex AI to accelerate AML compliance).

Step	What to test / tool
A/B experiments	Traffic splits, business metrics, confidence intervals - Vertex AI console
Scenario generation	Synthetic data, Monte Carlo and stress shocks - exhaustive scenario analysis
Backtesting & governance	Calibration, fairness audits, documentation and human checkpoints

“AI is not too big to fail - it is too complex to ignore.”

Start small: run controlled A/B arms, generate stressed scenarios (macro + micro shocks), backtest against historical crises, and surface early‑warning indicators so boards see resilience before a crisis arrives.

Start safely by treating Colombia's fast‑moving legal landscape as a design constraint, not an afterthought: map each pilot to the risk categories in the new AI Bill so high‑impact credit, AML or onboarding systems get extra controls, impact assessments and human checkpoints rather than a blind rollout (see the Baker McKenzie summary of Colombia AI Bill requirements and sanctions Baker McKenzie summary of Colombia AI Bill).

Pair that legal framing with CONPES 4144's playbook - use sandboxes, public–private testing and targeted upskilling to protect inclusion while iterating quickly (Colombia National AI Policy CONPES 4144 playbook).

Operational steps that limit risk: adopt internal AI management policies, classify systems by risk, require privacy‑by‑design and explainability for high‑risk flows, keep auditable documentation and human escalation paths, and run controlled A/B pilots before scaling.

For teams needing practical, workplace-ready skills - prompting, governance and safe deployment - consider role‑based training like the AI Essentials for Work syllabus (Nucamp) to build repeatable capacity without heavy engineering lift.

Bootcamp	Length	Cost (early / after)	Registration
AI Essentials for Work	15 Weeks	$3,582 / $3,942	Register for AI Essentials for Work (Nucamp)