All projects

Project 01 · AI · Essent

AI-first way of working, MCP servers & Skills.

Designing AI solutions end to end, from business problem to production, and making AI part of how the team refines and builds.

Client
Essent
Period
Aug 2026 – present
Role
AI Solutions Architect

The challenge

Many teams experiment with AI, yet few manage to turn it into a dependable part of their daily work. Without a shared approach, every developer uses AI in a different way and the quality of the output is hard to judge.

The goal was to move a development team to an AI-first way of working, with AI solutions that connect to internal systems and that can be tested before they reach production.

Approach

  1. Decomposing business problems
    Business problems broken down into AI solutions and agentic workflows that can be tested step by step.
  2. MCP servers and Agent Skills
    Connecting LLMs to internal systems and team knowledge, so AI works with the information the team already has.
  3. A shared AI way of working
    One approach for refinement and development with AI across the whole team.
  4. Evals, guardrails and observability
    Practices that make LLM-based features testable and keep them under control in production.

Key results

  • AI solutions designed from business problem to production, with every step testable
  • LLMs connected to internal systems and team knowledge through MCP servers and Agent Skills
  • One shared AI way of working for refinement and development, with evals, guardrails and observability built in

Technology

  • Claude
  • Model Context Protocol (MCP)
  • Agent Skills
  • TypeScript
  • Python
  • AWS