We have built software and data platforms for clients for more than twenty years, and for most of that time the industry has charged for hours. We would rather be paid for software that works. AI coding tools are the first thing that makes that a realistic offer, but only when the process around them is strict, and that process is what this post is about.
See all of Lineate’s technology partners
What changes for our team
A year ago, a few of our engineers were trying Claude on their own. Today it is part of their daily work on client projects, and the engineers who use it took the time to get certified.
The habits came before the certificates. Claude writes a plan before it writes any code, an engineer approves that plan, and every change is recorded in the repository. Those rules are why we are comfortable using AI on production systems that belong to our clients.
For the company, certification backs a decision we had already made: to sell working software rather than time. Handing routine coding to Claude is what makes that possible, and the review process is what keeps it safe.
What changes for clients
A Lineate team on a Claude-enabled project is small and senior, usually two engineers, each running several Claude agents in parallel. The engineers spend their time on design, architecture and review, and Claude takes the routine implementation work.
An engineer signs off on the implementation plan before any code is written and on the pull request before anything is merged. Plans, test plans and release notes are committed to version control alongside the code, so months later you can still see why a change was made. We judge the work by what ships.
How we use Claude on client work
The four projects below look very different, but they all follow the same rule: Claude drafts, and an engineer approves.
Claude Code in a HIPAA-regulated environment
For one of our HealthTech customers, Claude Code is the primary daily engineering tool across a HIPAA-regulated platform of seven repositories, built on .NET 8 and React 19. Engineers describe a change in plain language and Claude produces the technical design document for senior review. It runs code reviews that reason across all seven repositories at once, structured HIPAA gap analysis with locations and concrete fixes, and production release runbooks. Architecture docs, compliance reviews and runbooks are now standard artifacts of delivery, not separate workstreams.
Read the HealthTech case study
One AI workflow across two dozen applications
For a publisher with more than two dozen web and mobile applications in Java, PHP, React, Vue and server-rendered HTML, a stack rewrite was never going to be funded. So we standardized the delivery process instead. For any non-trivial change, Claude writes an implementation plan covering what changes, the files affected and a rollback path, and a senior engineer approves it before code is written. The same workflow holds in every codebase, and engineers work productively in parts of the portfolio they don’t know well.
Read the publishing case study
A repeatable Claude Code framework for data engineering
One of our data engineers turned his setup into an engineering problem. He built a lightweight framework inside the project repository: a CLAUDE.md entry point, playbooks for whole workflows such as deployments, and skills for narrow tasks. Git worktrees let Claude work several tickets in parallel, with human review at each milestone. Claude also investigates production incidents by querying databases and comparing data across jobs. A template version is now being unified for team-wide use.
Read the data engineering case study
AI agents in 24/7 support and monitoring
Lineate runs a 24/7 Tier 3 support, monitoring and application response team. Over the past year, we’ve supplemented it with frontier agents from Anthropic, OpenAI and Google, and we’re now packaging that capability for customers. A dedicated engineer learns each customer’s environment and tunes four agents: detection, evaluation, resolution and runbook validation. The result is faster detection, fewer false alarms and richer handoffs when a human is needed.
What’s next
We want more of our engineers certified, so Claude training is open to the whole engineering team. Beyond that, we are working on four things:
- Turning the playbooks, skills and worktrees setup from our data engineering project into a template every team can use. It started as one engineer’s system.
- Testing new uses of Claude. Each one stays an experiment until a senior engineer decides it is ready for daily work. At the moment that means plain-language tests that bring legacy code under coverage, turning requirements into user stories a team can build from, and documentation generated from the codebase for an engineering lead to sign off.
- Offering our AI-assisted 24/7 monitoring as a service, with a dedicated engineer and agents tuned to each customer’s environment.
- Keeping the rule that got us here. Our engineers set the direction, make the architecture decisions and review the code before it ships.
What won’t change: the goal stays the same — deliver working products, not hours.
If you want to see how this would work on your project, get in touch using the form below.
Frequently asked questions
Is Lineate a Claude partner?
Yes. Lineate is a Certified Services Partner in the Claude Partner Network, and our engineers hold Claude certifications.
How does Lineate use Claude on client projects?
Our engineers use Claude Code to write implementation plans, design documents, code reviews and release runbooks. An engineer approves each plan before any code is written and reviews the pull request before it ships.
Can Claude Code be used on a HIPAA-regulated platform?
Yes. For one HealthTech client, Claude Code is the primary daily engineering tool across a HIPAA-regulated platform of seven repositories. It also produces structured HIPAA gap analysis, and senior engineers review what it produces.
How big is a Lineate team on a Claude project?
Usually two senior engineers, each running several Claude agents in parallel.
