How long does it take to develop a web application?

Timelines vary based on complexity. An MVP can be ready in 4-8 weeks, while a full enterprise application may take 3-6 months. We use agile methodologies to deliver value from the very first weeks.

What technologies do you use for development?

We use modern technologies including React, Next.js, TypeScript, Node.js, Python, React Native, and Flutter. For AI, we work with OpenAI, LangChain, and custom models. Our infrastructure is powered by AWS, Vercel, and Google Cloud.

Do you work with companies in the United States?

Yes, we work with clients in over 5 countries including the United States, Bolivia, Chile, Peru, and Argentina. Our team is fully prepared to work remotely with constant communication, aligned with US time zones.

How much does a software development project cost?

Each project is unique and cost depends on scope, complexity, and required technologies. We offer free consultations where we evaluate your project and provide a detailed estimate. Contact us to schedule a meeting.

Do you offer post-launch maintenance and support?

Yes, we offer maintenance and post-launch support plans that include bug fixes, security updates, performance optimization, and new features. We ensure your product continues to work flawlessly.

What is nearshore software development?

Nearshore development means working with a team in a similar time zone. Based in Bolivia (GMT-4), we overlap with US business hours, enabling real-time collaboration at competitive rates without the communication challenges of offshore teams.

BORING

, but not that Boring

Back

CASE STUDY

Semapa

Sector

AI-powered management system · +120,000 clients · WhatsApp complaints

Client

SEMAPA · Municipal Drinking Water Service · Cochabamba, Bolivia

Internal management platform for SEMAPA that centralizes the records of more than 120,000 clients, WhatsApp complaints and technical and commercial reports connected to the central system — all monitored in real time through AI agents and machine learning.

Scroll to explore

AI Agent✦Machine Learning✦WhatsApp✦Internal management✦Utility✦Cochabamba✦AI Agent✦Machine Learning✦WhatsApp✦Internal management✦Utility✦Cochabamba✦AI Agent✦Machine Learning✦WhatsApp✦Internal management✦Utility✦Cochabamba✦AI Agent✦Machine Learning✦WhatsApp✦Internal management✦Utility✦Cochabamba✦

About

For SEMAPA (Cochabamba's water utility) we built a comprehensive AI management system that orchestrates the full customer service operation and the technical/commercial operation behind the company. The platform records more than 120,000 active clients, receives and classifies complaints coming mainly from WhatsApp (also Web, Facebook, Instagram and ManyChat), generates technical and commercial reports connected directly to the company's central system and gives management full visibility — with heatmaps of complaints in Cochabamba (169,211 historical complaints), a real-time conversation inbox with AI Agent + Human Agent, management of scheduled water-cut zones with polygons over the map, exportable logbook, configurable AI model (Gemini Flash latest as primary and GPT-4o mini as automatic fallback), knowledge base, editable system prompts, rules, tools, incidents and services. The system lets you see absolutely everything happening behind the company — from an individual georeferenced complaint (T-317253 · Ana Lopez · calle Ayacucho 300 · In Progress) to the detail of pending invoices that the AI agent answers on WhatsApp with a month-by-month breakdown.

Project brief

From complexity to a working product.

The challenge

SEMAPA needed to modernize customer service and the internal operation behind a base of more than 120,000 clients: complaints were coming through scattered channels (WhatsApp, web, social), technical and commercial reports lived in separate systems and management had no consolidated visibility of what was happening in the city in real time. We had to automate invoice and complaint inquiries without losing human control, connect everything to the existing central system and give administrators a single panel to see the pulse of the company.

The solution

We designed an AI management system in Next.js with its own login (semapa.bo) and a 14-module admin panel: Dashboard with weekly/monthly report and service channels (WhatsApp, Facebook, Instagram, Web, ManyChat), Complaint Management with a Cochabamba heatmap (169,211 historical complaints) and filters by type/status/origin (Manual/AI/App), Live Chat, Operator, WhatsApp with real-time inbox and AI/human control (Take control / AI Active), Tools, editable System Prompts, Knowledge Base, Rules, Users, Cut Locations with polygons over the map and a logbook exportable to PDF, Incidents, AI Model with a switch between Gemini Flash (primary) and GPT-4o mini (automatic fallback) applied in real time to the 3 agents, and Services. The AI agent answers invoice inquiries with a month-by-month breakdown (February Bs 207.70, March Bs 136.30, etc.) and hands off to a human operator outside working hours.

Platform

A tour of the system

Control Panel · Weekly report and channels

Dashboard with daily complaint metrics (technical / commercial), totals by status (In Progress, Resolved, Pending) and breakdown by origin: AI Agent Bot 257 (98%), Human Agent 4 (2%), Mobile App 0%. Service channels with WhatsApp 10 (4%), Web 251 (96%), Facebook, Instagram and ManyChat.

Complaint Management · Cochabamba heatmap

Heatmap with 50 technical and 50 commercial complaints georeferenced over Cochabamba, filters by type/status/origin (Manual/AI/App) and a side summary with tickets T-317253 Ana Lopez, T-317252 Carlos Vega, T-317251 Juan Perez Lopez — Showing 100 of 169,211 complaints.

WhatsApp · Real-time conversations

Inbox with 49 active conversations and an open chat with Arteaga G Rene (cod 024509400): the AI Agent responds with a month-by-month breakdown of pending invoices (February Bs 207.70, March Bs 136.30, April Bs 138.30, May Bs 140.00) and hands off to a human operator. CTAs AI History / AI Active / Take control.

Cut Locations · Polygons over the map

Recording and managing scheduled water-cut zones with polygons drawn over the map of Cochabamba (Molle ZS-002, chimba ZS-001, RED ALTA BARRRILETE) and a logbook exportable to PDF by date range.

AI Model · Gemini Flash + GPT-4o mini fallback

Configuration of the primary LLM with a real-time switch between Gemini Flash (latest, Vertex AI) as Primary and GPT-4o mini (OpenAI API) as Fallback. Automatic fallback enabled: if Gemini fails, the system uses OpenAI without interruption — change applied to the 3 agents in under 30s.

Highlights

Clients registered on the platform

Complaints handled historically

Service handled by AI Agent · 2% human

Conversations from the Web channel · rest WhatsApp

Back to all projects

Back

CASE STUDY

Semapa

Sector

AI-powered management system · +120,000 clients · WhatsApp complaints

Client

SEMAPA · Municipal Drinking Water Service · Cochabamba, Bolivia

Scroll to explore

About

Project brief

From complexity to a working product.

The challenge

The solution

Platform

A tour of the system

Control Panel · Weekly report and channels

Complaint Management · Cochabamba heatmap

WhatsApp · Real-time conversations

Cut Locations · Polygons over the map

Recording and managing scheduled water-cut zones with polygons drawn over the map of Cochabamba (Molle ZS-002, chimba ZS-001, RED ALTA BARRRILETE) and a logbook exportable to PDF by date range.

AI Model · Gemini Flash + GPT-4o mini fallback

Highlights

Clients registered on the platform

Complaints handled historically

Service handled by AI Agent · 2% human

Conversations from the Web channel · rest WhatsApp

Back to all projects