Portfolio

Work

Projects we delivered — challenges, solutions and results.

AI Voice · Bookings

AI Voice Handling for Equipment Reservation System

200

calls/day

call-centre staff

4 tyg.

deployment time

Context

The client — a construction equipment rental company — was handling hundreds of calls daily with a limited team. Queues, booking errors and no after-hours support generated losses and customer frustration.

Challenge

Automating phone-based reservations with real-time fleet management system integration, handling natural language and regional speech patterns.

Solution

We deployed a voice agent built on OpenAI Realtime API with Twilio integration. The agent conducts the conversation, checks equipment availability via API, creates the booking and sends an SMS confirmation. Runs 24/7, handles up to 200 simultaneous calls.

OpenAI Realtime APITwilioCloudflare Workers

AI Sales · Lead

AI Lead Assistant & Initial Quotes

3 min

response time

4×

more qualified leads

60%

less sales work

Context

A service agency was losing potential clients due to slow response times on quote requests. Sales reps spent hours qualifying leads, most of which were a poor fit. Conversion from inquiry to meeting was below 10%.

Challenge

Building an agent that understands the company's service specifics, asks the right qualifying questions and generates a sensible quote — without hallucinations and without involving sales at the early stage.

Solution

GPT-4o-based agent fine-tuned on the company's historical quotes. HubSpot integration for deal creation and Calendly for meeting booking. Workflow orchestrated via n8n with human escalation rules when the conversation requires negotiation.

GPT-4on8nHubSpot APICalendly

MCP · Onboarding

MCP Server for New Platform Client Onboarding

45 min

full onboarding

support tickets

100%

self-service

Context

A B2B SaaS with a growing client base had a bottleneck: onboarding a new client required 2–3 hours of engineer work and was a scaling constraint. New clients waited up to 5 business days to get their environment running.

Challenge

Integrating 6 internal systems (CRM, Confluence, GitHub, Kubernetes, Vault, monitoring) into a single coherent context accessible to an AI agent without violating security policies.

Solution

MCP server with per-client granular access control. Claude agent walks the client through an onboarding checklist: creates a Kubernetes tenant, configures Vault secrets, creates a repository, sends credentials and verifies configuration correctness. Every step is auditable.

MCP ProtocolClaude APIKubernetesTerraformVault

Have a similar project?

Book a free consultation — tell us about your challenge.

Book a free consultation