How Althea is building an AI-native engineering model with Andela & Anthropic

‍SNAPSHOT

Customer: Althea
Industry: Insurtech
Use case: AI-native software development and context-aware engineering
Tools: Claude Code, Notion, Compound Engineering

What was built: An AI-native engineering workflow that uses Claude Code across development, review, quality assurance, and technical decision-making. The workflow connects the product and architecture context developers need to make better decisions, and it gives engineers ownership of a much wider part of the product lifecycle.

Early impact:

• 100% Claude Code adoption across the company
• Faster engineering workflows with less back-and-forth during QA
• Higher-quality development through AI-assisted code review and security checks
• Better continuity of context across product decisions, architecture, documentation, and engineering work
• Engineering ownership of requirements, product decisions, prototyping, and iteration, with less handoff to a dedicated product management function

Problem: As AI takes on more engineering work, context becomes critical

Althea was built in an environment where AI is already part of how work gets done. Claude Code supports software development from planning and implementation through code review and quality assurance.

As the team pushed further into AI-native engineering, a more fundamental challenge emerged: AI is only as effective as the context it can access. Product requirements evolve, architectural decisions change — and important context lives across pull requests, meetings, tickets, and documentation. When that information is fragmented or outdated, AI can work from incomplete assumptions.

For Althea, the opportunity became bigger than accelerating individual tasks. It became about creating an engineering environment where AI could operate with a reliable understanding of the product and technical decisions behind the work.

The goal is an engineering model where AI does more of the execution without losing the context, judgment, and quality controls required to build reliable software.

Context lives everywhere

At Althea, an important decision might surface in a pull request, meeting, ticket, or architecture document.

As AI participates across more of the development lifecycle, keeping that context aligned becomes critical. The team needed a way to preserve decisions as the product evolved, so each new piece of work could build on what had already been decided rather than starting from an isolated prompt.

Solution: Building an AI-native engineering workflow with Claude Code

Andela AI Engineers work directly within Althea’s development environment, helping extend an engineering model where AI is already integrated throughout the development lifecycle.

Claude Code supports work from architecture and planning through implementation, review, testing, and security checks. The team adapts how it uses Claude Code based on the complexity of the task, while Compound Engineering, an open-source prompt framework, helps make repeatable workflows easier to apply across research, implementation, and testing.

"Claude Code hasn’t just made our engineers faster. It’s changing how we organize the work. Engineers can own product outcomes end to end, from gathering context and defining requirements through implementation and verification. For a team our size, that fundamentally changes the scope and ambition of what we can build.”
‍
— Geoffroy Bablon, CEO at Althea

The Claude Code engineering playbook

• Keep product context connected to engineering work: Product decisions are captured in shared documentation so developers and AI systems can reference the current state of the product rather than relying on isolated tickets or conversations.
• Make architecture decisions persistent: When new engineering work conflicts with or changes an existing architecture decision, the workflow can surface the discrepancy so documentation stays aligned with the system being built.
• Use Claude Code across the development lifecycle: Engineers use Claude Code from planning through implementation, using Plan Mode to reason through the technical approach and specialized subagents to support code review, testing, and security analysis before work moves forward.
• Capture decisions where they happen: Meeting transcription and documentation help preserve decisions made outside the codebase, reducing the likelihood that important product context disappears between conversations and implementation.
• Standardize repeatable engineering tasks: Compound Engineering gives the team reusable workflows within Claude Code for researching a problem, understanding the existing system, writing code, and testing the result, with specialized subagents used where tasks benefit from independent analysis or parallel execution.
• Keep human judgment at the center: AI accelerates execution, but architectural decisions, compliance questions, and technical judgment remain with the team. Althea also maintains access to legal expertise when questions touch the regulatory environment.

Impact: Faster development without sacrificing quality or context

The clearest impact is speed without a tradeoff in quality. Claude Code helps Althea’s team not only implement code, but also review it and perform security checks before work reaches later stages of QA.

That has helped shorten feedback loops while giving the team better continuity across product and architecture decisions.

The early results:

• 100% adoption of Claude Code across Althea
• Shorter QA feedback loops through AI-assisted implementation, review, and security checks
• Improved engineering velocity and quality
• Better continuity of context across product decisions, architecture, and implementation
• AI experimentation beyond engineering, with employees across functions building their own tools and workflows

The bigger shift is in the work itself

As AI takes on more execution, engineers can spend more time evaluating architecture, resolving tradeoffs, validating outputs, and applying technical judgment. For Althea, Claude Code is becoming part of an operating model built around persistent context, continuous experimentation, and human judgment.

“Althea is showing what AI-native engineering can look like when Claude Code and the Claude API are embedded throughout the development lifecycle. By building workflows that preserve product and technical context, the team can use AI across more of the work while keeping human judgment at the center. That combination is helping create a faster, higher-quality, and increasingly AI-native way to build software at Althea.”
‍
– Kennith Jackson, SVP, AI Solutions, Andela