AI Assisted Software Engineering - Scale with GitHub Copilot

Our Proposal

It is based on the rapid adoption of AI in product development, leveraging current Azure capabilities and modern AI tools such as GitHub Copilot, along with Baufest’s expertise in software development and applied AI solutions.

Our proposal is structured into two phases, described below:

Create Phase

We design and define the rules for building your product using generative AI—covering both the technical infrastructure and architecture, as well as the product’s functional definition and how users intend to interact with it.

This phase is divided into two stages:

Technical Setup

We establish the technical capabilities required for development, including:

• Creation of development environments and CI/CD support
• Solution architecture, including technical features such as security, observability, persistence, etc.

Functional Setup

We integrate all product-related information into the AI to enable its development. This includes:

• UI design and navigation (implementing the brand guidelines)

• Granting AI access to product documentation to:

  • Define user stories
  • Establish business rules
  • Specify data flows
  • Test deliverables against functional requirements and business rules

Completion of the Create Phase

At the end of this phase, we build an MVP workflow that enables the implementation of a user story fully powered by AI across all stages of the development lifecycle (refinement, design, development, testing, and deployment). This results in the first user story developed using the new working methodology.

Operate Phase We operate the AI-enabled development environment to build all product functionalities, evolve the business, and integrate with the client’s existing software assets. Through agile Scrum sprints, we iterate during this phase—releasing new capabilities and enhancing existing ones—while aligning with the company’s digital strategy.

Work Plan

Technical Setup (1 to 3 months)

• Establish the development environment including all baseline artifacts for the CI/CD lifecycle
• Implement architecture using technologies, languages, and tools aligned with the client’s development culture
• Deliver technical enablers to support the MVP: Security (authentication and authorization), Observability and monitoring, Data persistence and inter-layer communication and Responsive navigation

Functional Setup (2 to 8 weeks)

• Implement brand guidelines by connecting AI to style definitions (colors, fonts, sizes)
• Integrate AI with repositories containing: Initial product definition, Business rules, Test cases, Approval workflows, Production support systems

Note: The exact duration of the Create phase will be estimated at the beginning of the project, based on the characteristics of the software to be developed and the current capabilities of each client’s IT department. Baufest will provide a team sized according to the backlog and required development velocity, responsible for building and evolving the product—covering business features, technical capabilities, integrations, analytics, and any feature that adds value and enables growth.

Key Differentiators of Our Proposal

The key to successful AI adoption is ensuring consistency in product documentation and information used to feed LLMs, enabling them to generate accurate and valuable outputs.

At Baufest, we use AI to validate the correctness and completeness of all product-related repositories.

Example: A documented business rule may state that a tax must be withheld at the point of sale, but the formula to calculate it is missing. Ensuring consistency is critical to generating correct code.

We feed software development agents with precise and optimized functional and technical information.

Agent Strategy Tailored to Each Task

Our strategy involves using different AI techniques to address the specific challenges of each stage in software development.

• We use Vibe Coding for creative exploration tasks, where the goal is to introduce new concepts and explore possibilities that enrich product design.
• We apply deterministic workflows for precise implementation of requirements, ensuring compliance with business rules, security standards, and defined architecture.

We base development on agent workflow models using the capabilities of Microsoft’s Azure AI Foundry platform, which we extend, customize, and integrate with third-party tools through the Microsoft Agent Framework.