AI for banking in Paraguay: what to automate

Ready for a pilot?

Tick what you already have; heavier items stall a pilot more. Nothing is sent.

Pilot readiness for a bank

0 % · 0 of 6 controls that applyFirst missing: process measured and compliance sign-off

Not yet: the basics for testing with real data are missing.

Higher-risk controls weigh more in the score.

We know how many files or complaints arrive per month and how long each takes.
Compliance and risk approved what data the AI sees and what a person decides.
There is a read-only path to the core or a copy, without touching production.
We have 100 real anonymized cases to test before going live.
Every action is logged with date, user and version, as evidence for the SIB.
A named manager is accountable for the pilot.

What it costs and when it pays back

Hours freed by AI minus what it costs to run. Examples are marked; public prices link to their source.

Files or complaints per month1,000

Minutes of work per file today75 min

Minutes per file with AI (example)15 min

AI model

Cost per hour (PYG)

Floor: the minimum-wage hourly value (MTESS, July 2026). Typical real cost (example): ×2.5, a salary of two minimum wages plus employer contributions and the 13th month. Source

Today1,250 h/month

Automated250 h/month

Value of hours freed per month1,000 h × PYG 12,683: PYG 12,683,000
AI model public price1,000 files × USD 0.0040 (example: 30,000 input + 2,000 output tokens; gpt-6-luna at USD 0.10 / 0.50 per million) · OpenAI · API pricing: −PYG 22,778
Platform, monitoring and support exampleTypical monthly range: −PYG 1,708,341 to −PYG 4,555,576
Net savings per month: PYG 8,104,646 to PYG 10,951,881
Core read access, compliance rules, testing and pilot typical range · exampleOne-off, up front: PYG 45,555,760 to PYG 113,889,400
Months to pay back the investment: 4.2 to 14 months

Tokens: the unit AI providers bill for the text the model reads (input) and writes (output). OpenAI puts 100 tokens at about 75 English words.

Range: the lower figure uses low costs, the higher one high costs. Public prices checked on 11 Oct 2026. USD prices converted at PYG 5,694.47 per dollar (reference rate of 9 Oct 2026): BCP

estimated

4-week pilot

A typical plan; real timing depends on your systems.

Pilot · weeks 1 to 4

Week 1Pick and measureOne process (KYC file or complaints) and its baseline of time and errors.KPI: Signed baseline
Week 2Rules and accessCompliance defines what the AI sees; read-only access and every step logged.KPI: Access approved
Week 3Test on 100 casesAI assembles files on real anonymized cases; the team corrects.KPI: Correct fields (example target: 95%)
Week 4Scoped real useOne team uses the flow and it is compared to the baseline.KPI: Minutes per file vs. baseline

Before you start

Process chosen and monthly volume measured
Written compliance rules: what the AI sees and what a person decides
Read-only access to the core or a copy
100 real anonymized cases for testing
Every action logged with user, date and version
Pilot owner and success metric agreed

Frequently asked questions

How do you handle security and customer data?

Role-based access, encryption, audit logs and deployments on controlled infrastructure. The security model is defined with you before writing code, aligned with your policies and regulator requirements.

Does it integrate with our banking core?

Yes. Integration with core and legacy systems is our specialty: we assess your core in the diagnostic and define the safest path, without touching what already works.

Can AI execute sensitive operations?

Nothing critical is delegated without a safety net: bounded action scope, automated validations and human authorization on sensitive operations. AI proposes, the system validates, the human authorizes.

Where should we start?

Where payback is fastest and risk lowest: typically customer service or document processing. We define it in the diagnostic with your numbers.