AI in Delivery: What We’ve Built Over the Last Six Months

Our CEO and Co-founder, Anton Morozov, recently spoke at Growth Factory about a topic that has become one of the biggest challenges for today’s IT businesses: “AI in Delivery: What We’ve Built Over the Last Six Months.”

Below are the key takeaways from his presentation, highlighting the AI solutions that are already running in production.

What We’ve Built

AI Agentic Loop

An end-to-end development workflow where AI agents:

• refine backlog items,
• ask clarifying questions in Slack,
• generate code,
• and prepare Merge Requests.

AI-Powered Log Monitoring

Our large-scale cloud environment is now monitored by an AI agent that continuously analyzes logs and automatically detects anomalies before they become critical issues.

Architecture Audit

We developed an architecture auditing plugin powered by a network of specialized sub-agents. It produces a comprehensive assessment of any project, evaluating code quality, architecture, security, and compliance with internal engineering standards.

Back-Office Automation (n8n)

We’ve introduced AI agents to automate operational workflows. For example, every morning our internal bot delivers reports on team utilization and highlights potential financial risks, allowing managers to react proactively.

AI-Assisted Project Estimation

We developed a custom Claude skill implementing our 3-point estimation methodology. It helps teams produce estimates faster, more consistently, and at a lower cost.

AI-Powered Code Review

We built wl-mr-review, a plugin that streamlines the code review process.

A Tech Lead submits a completed Merge Request to Claude. The plugin connects to the GitLab API through an external MCP server and evaluates the Merge Request against predefined code quality standards. The Tech Lead reviews the generated feedback, removes irrelevant comments, refines the remaining ones, and publishes them directly in GitLab.

As a result, code reviews are performed locally, faster, and without unnecessary delays.

Human in the Loop

The core principle behind all of our AI initiatives is Human in the Loop.

AI has significantly accelerated software delivery, but the final responsibility always remains with people. Architectural decisions, Merge Request approvals, and business requirements validation continue to be performed by experienced engineers.

At WebbyLab, AI doesn’t replace expertise – it amplifies it by automating repetitive work and allowing specialists to focus on solving complex engineering problems.