How is N-AI consulting priced?

Each project is quoted by scope, technical depth and duration. There are three models: fixed-scope project (from idea to production), monthly engagement or hourly advisory.

How does a project start?

With a no-commitment introductory conversation to understand the problem and the available data. It leads to a proposal with scope, engagement model and plan.

What makes N-AI different?

Every project is led directly by the founder, Nancy Rodríguez. She combines the quantitative method of an economist and statistician with hands-on expertise in data science, machine learning and data architecture.

What services does N-AI offer?

Artificial intelligence and machine learning, intelligent automation, advanced analytics and BI, and data architecture and governance.

Who does AI consulting in Costa Rica?

N-AI (n-ai.dev), founded by Nancy Rodríguez, provides AI, automation, advanced analytics and data architecture consulting for companies and institutions across Costa Rica and Latin America.

Who does public procurement intelligence in Costa Rica?

N-AI built the intelligence architecture for SICOP at Costa Rica's Comptroller General (Contraloría General de la República): anomaly detection across 47 patterns, risk scoring across 8 institutional dimensions and preventive monitoring.

Back to home

Case study

Public procurement intelligence for Costa Rica's Comptroller General

Data architecture and artificial intelligence to move SICOP oversight from reactive to preventive.

Records analyzed: >2.4M
Anomaly patterns: 47
institutional risk dimensions: 8
time to compute a risk score: <1 s

Summary

Context

Public procurement oversight was reactive: irregularities surfaced after the damage was done, with data spread across different sources and taxonomies.

Outcome

Risk becomes visible before it turns into a legal case: auditors act on signals, not on findings.

The challenge

Why oversight arrived too late

Data spread across different sources and formats
Classifications that change between institutions
Irregularities detected after the damage was done
Systemic risk patterns nobody modeled

Sources and architecture

Seven sources, seven stages

Sources

Solicitations
Awards
Contracts
Addenda
Supplier registry
Sanctions
Institutional data

Pipeline stages

1
Extraction
Continuously pulls records from every SICOP source.
2
Inventory
Documents every field and every relationship between sources.
3
Unification
Brings the different formats together into a single data model.
4
Validation
Checks types, consistency and traceability before the data is used.
5
Anomaly detection
Evaluates 47 patterns across time, structure and behavior signals.
6
Risk scoring
Combines 8 dimensions into an auditable score per institution.
7
Preventive monitoring
Shows events, trends and peer comparisons to decision makers.

AI layer

47 patterns and 8 risk dimensions

Rules catch the obvious and models catch the subtle: unusual award velocity, supplier concentration, out-of-range prices and addendum sequences that often precede irregularities. Each institution gets a score that breaks down into 8 dimensions.

8-dimension compositeComposite 62/100

Contracting velocity78
Supplier concentration62
Price dispersion71
Addendum exposure55
Sanction proximity34
Transparency posture82
Control maturity49
Signal density67

Institutions

12 institutions, anonymized

Risk score for each institution, on a 0 to 100 scale.

low · 3 · medium · 4 · high · 3 · critical · 2

low
Institution A
23
low
Institution B
31
low
Institution C
28
medium
Institution D
47
medium
Institution E
52
medium
Institution F
58
medium
Institution G
49
high
Institution H
71
high
Institution I
76
high
Institution J
68
critical
Institution K
84
critical
Institution L
91

Project results with anonymized institutions.

Impact

What changes for decision makers

Preventive oversight

Risk is visible before it turns into a legal case: action is taken on signals, not findings.

Faster decisions

Decision makers see risk on the first screen and reach a defensible decision in minutes, not weeks.

Verifiable transparency

Every score traces back to its components, and every component to a record.

Facing a similar challenge?

An introductory conversation identifies how to apply this approach to your data.

Book a diagnostic