Case studies

Measurable results. Real, anonymized cases.

For confidentiality, some clients prefer not to appear. We keep the technical details and the metrics; we share the name in a meeting if appropriate.

Featured case
SPANISH FINTECH · 80 EMPLOYEES · 2026

Contract-review assistant over 12,000 legal documents.

Context

A Series-A fintech with a 3-person legal team reviewing about 400 contracts a month across institutional clients and partners. Average time per contract: 45 minutes. Chronic backlog of 2–3 weeks.

Challenge

Reduce review time without sacrificing precision on critical clauses (SLAs, liability, intellectual property), keeping all data inside the client's VPC — it couldn't go to external providers.

Solution

A review assistant with Claude Sonnet served via AWS Bedrock (data in the EU), RAG over the historical base of 12,000 contracts with embeddings in pgvector, and a human approval flow that keeps the lawyer in the loop for every recommendation.

Claude SonnetAWS BedrockpgvectorNext.jsPython FastAPITerraform
Review time
−73%
from 45 min to 12 min per contract
Critical-clause recall
97%
validated on a 400-contract test
Backlog
0
vs. 2–3 weeks before
Time to production
9 wks
POC + MVP + rollout

“We went from being the company's bottleneck to the team leading internal AI adoption. VEZGALABS understood our compliance constraints before our own DPO did.”

— Head of Legal, anonymized fintech
Other cases

More recent work.

Legaltech · 40 employees

Automatic classification of court rulings

−62%
cataloguing time
LLMEmbeddingsPG Vector
Retail · 250 employees

Conversational assistant for the store team

+18%
post-sale NPS
ClaudeRAGWhatsApp API
Industrial · 120 employees

Anomaly detection on production lines

−47%
unplanned downtime
PythonPyTorchGrafana
Healthtech · 60 employees

Structured extraction from medical reports

94%
accuracy vs. human
GPT-4oJSON schemaFHIR
Free assessment · 30 min

Want to talk about your use case?

We help you identify where AI delivers real ROI in your company. No strings attached, with an actionable roadmap by the end of the call.

Book an assessment See case studies