Prudence

Your AI Analyst.

Upload your data. Prudence builds your first dashboard. Then work with an AI analyst that understands your business.

Prudence turns enterprise data into continuously evolving intelligence. Upload spreadsheets, connect databases, generate dashboards automatically, and refine them through natural language.

Dashboard-first workspaceAnalyst terminal, not chatbot UIOrganizational knowledge built with OKFCloud and on-prem deployment paths
prudence://workspace/credit-risk

Credit Risk Overview

Portfolio dashboard generated from uploaded loan data

Generated dashboardTraceable SQLPackage-aware metrics

Last run

14s • 8 widgets updated

Total Exposure

$184.2M+6.2%

NPL Ratio

8.7%+1.1%

PAR 30+

12.4%+0.8%

Flagged Branches

142 new

NPL Trend

Generated from monthly delinquency history

last 12 months

Branch Exposure

Central84%
North68%
East57%
South41%

Top alert

Lahore North branch shows a 2σ increase in PAR 30+ after April.

Analyst Terminal

The terminal modifies the dashboard, not just a chat transcript.

run #0182

> show NPL trend by branch and add a risky branches export

✓ Reading schema

✓ Planning analysis

✓ Running SQL

✓ Updating dashboard

✓ Saving OKF note

✓ Preparing CSV export

Works with

CSVExcelParquetPostgresBigQuerySnowflakeSQL Server

A new interface for enterprise analytics.

Prudence moves the analyst loop from disconnected spreadsheets and BI tabs into one workspace that keeps context between runs.

01

Connect data

Start with uploads or point Prudence at the systems your team already uses.

CSV, Excel, Parquet
Postgres, BigQuery, Snowflake
Private and cloud data sources

02

Prudence understands your data

The workspace profiles schema, detects domain context, and builds semantic understanding before analysis starts.

Schema profiling
Domain detection
Semantic column hints

03

Generates dashboards

Prudence builds the first useful dashboard immediately so the user starts from analysis, not setup.

Portfolio overview
KPIs, charts, and tables
Package-aware defaults

04

Work with your AI analyst

Refine the dashboard through a terminal-style interface that can reason with natural language, SQL, and Python.

Terminal workflow
Dashboard edits
Traceable execution steps

05

Build organizational memory

Insights become reusable knowledge so the next analysis starts with more context than the last one.

OKF knowledge base
Consistent definitions
Reusable reports and notes

The analyst workspace is the product.

The dashboard is the primary artifact. The terminal is how the user changes it. Prudence keeps data, dashboards, and knowledge in one working environment.

prudence://workspace/credit-risk

Credit Risk Overview

Portfolio dashboard generated from uploaded loan data

Generated dashboardTraceable SQLPackage-aware metrics

Last run

14s • 8 widgets updated

Total Exposure

$184.2M+6.2%

NPL Ratio

8.7%+1.1%

PAR 30+

12.4%+0.8%

Flagged Branches

142 new

NPL Trend

Generated from monthly delinquency history

last 12 months

Branch Exposure

Central84%
North68%
East57%
South41%

Top alert

Lahore North branch shows a 2σ increase in PAR 30+ after April.

Analyst Terminal

The terminal modifies the dashboard, not just a chat transcript.

run #0182

> show NPL trend by branch and add a risky branches export

✓ Reading schema

✓ Planning analysis

✓ Running SQL

✓ Updating dashboard

✓ Saving OKF note

✓ Preparing CSV export

Dashboard

Start from a generated dashboard instead of a blank canvas or a blank prompt.

Analyst Terminal

Ask for edits, breakdowns, exports, and new views with the execution path visible to the user.

Knowledge

Save insights, definitions, and assumptions into a persistent knowledge layer that compounds over time.

Data Explorer

Inspect profiles, columns, and table structure directly inside the workspace before you ask bigger questions.

Reports

Export charts, tables, dashboard artifacts, and narrative summaries without rebuilding the work elsewhere.

Packages and skills give Prudence context.

Packages teach Prudence how to think about your business. Skills let Prudence analyze data the way experienced analysts do.

Packages

Domain context, prebuilt metrics, and better defaults

v0.1

Credit Risk

Delinquency, NPL, PAR, branch and product exposure analysis.

Inventory Risk

Stock velocity, overhang, replenishment, and margin exposure.

Finance

P&L drilldowns, management reporting, variance analysis, and controls.

Operations

Service performance, throughput, SLA exceptions, and process bottlenecks.

Telecom

Wallet behavior, lending performance, repayment trends, and cohort movement.

Fraud

Anomaly detection, case review queues, and investigative analysis workflows.

Skills

Analyst methods Prudence can bring to the workspace

SQLPythonExcelStatisticsVisualizationForecastingCredit RiskInvestigations

Why it matters

Skills help Prudence analyze data the way experienced analysts do. Packages keep that analysis grounded in the domain logic that teams already care about.

Knowledge that survives the report cycle.

Prudence continuously builds an organizational knowledge base. Insights are never lost, reports become reusable, and business definitions stay consistent.

Every analysis can become a reusable note, assumption, or definition.
Definitions stay close to the dashboard and the data that produced them.
The next analyst session starts with context instead of starting over.

okf/portfolio/npl-trends.md

---

type: risk_insight

confidence: medium

source_dataset: loan_listing_may2026.csv

last_verified: 2026-06-30

related:

- portfolio/branch-risk.md

- portfolio/npl-trends.md

---

 

# NPL Trend

 

Summary: NPL ratio rose from 7.1% to 8.7% between December and May.

Interpretation: Concentrated in retail-heavy branches with weaker recoveries.

Why Prudence.

The difference is continuity. Prudence keeps the dashboard, the context, and the knowledge from the last run so the next cycle starts ahead of the previous one.

Traditional

Monthly work starts from scratch

1Open spreadsheet
2Write SQL
3Create dashboard
4Export report
5Repeat next month

Prudence

Each run begins with more context than the last one

1Upload data
2Dashboard generated
3Ask questions
4Refine analysis
5Knowledge grows

Designed for real enterprise environments.

Prudence is built for teams that need deployment control as much as analytical speed. Start on GCP, move on-prem later, and keep the architecture portable.

Cloud

Move quickly with a managed deployment path.

GCP
Cloud Run
Cloud SQL
GCS
Memorystore

On-prem

Run close to regulated data and internal infrastructure.

Docker or Kubernetes
Postgres
MinIO
Redis
Local model provider

Portability

The core architecture stays replaceable.

Private AIBring your own modelsBring your own storageDuckDB computeMarkdown knowledgeOpen formatsNo vendor lock-in

Deploy on GCP first, move on-prem later, and keep the data, knowledge, and compute layers under your control the whole way.

Built on simple, durable technologies.

Prudence avoids unnecessary infrastructure complexity. The stack stays inspectable, portable, and easy to reason about as the product grows.

DuckDB

Analytical core for per-workspace compute and fast local execution.

Markdown

Human-readable knowledge artifacts that remain portable across deployments.

Open Knowledge Format

Structured notes, assumptions, and reusable findings that compound over time.

Terraform

GCP-first infrastructure defined cleanly enough to support on-prem evolution.

Open APIs

Frontend, backend, and agent layers stay decoupled behind stable interfaces.

Portable deployments

Cloud today, private infrastructure tomorrow, without rewriting core logic.

Pricing during the pilot phase.

Prudence is currently rolled out through scoped pilots and enterprise deployments. Pricing depends on data environment, package mix, and deployment requirements.

Pilot Workspace

For teams validating one core workflow with one package.

Custom
Initial workspace setup
Guided onboarding
Shared deployment planning

Team Rollout

For analysts who need multiple dashboards, exports, and package coverage.

Custom
Multiple analysts
Expanded package scope
Reporting and workflow support

Enterprise

For regulated environments that need private AI and infrastructure control.

Custom
On-prem options
Private model support
Security and architecture review

Availability

Prudence is currently available through pilot rollouts and enterprise design partnerships.

Request Demo

Final CTA

Stop spending time building reports.

Start working with an AI analyst that understands your business and gets better with every analysis.

Request Demo