Our new team structure: the Builder POD

We've moved away from traditional scrum teams toward small, outcome-focused units we call Builder PODs. Each POD is assembled around a specific initiative and pairs one or two Forward Deployed Engineers (FDEs) with Context Owners drawn from product, customer success, and leadership.

Context Owners: bring deep knowledge of the problem we're solving, the customer we're solving it for, and have foundational skills in product management or design.

Forward Deployed Engineers: bring the ability to turn that understanding into working software fast, and over time, develop the same deep knowledge of the problem and customer as the Context Owners.

Behind the PODs sit our Platform Engineers, who maintain the shared infrastructure, tooling, AI tooling, and general platform standards that every POD relies on. 

This allows the PODs to move quickly without each one reinventing the foundations. They rapidly ship products that solve customer needs, while the platform team keeps the ground solid underneath them. Our FDE’s and Context Owners, are constantly working horizontally with Platform Engineers to ensure the highest leverage capabilities are developed.

‍
Discovery: Prototyping with customers, in high fidelity

The biggest change within our new process is when the customer enters the loop. Instead of gathering requirements, disappearing for weeks, and returning with a demo, our Context Owners and FDEs build prototypes directly with customers at the very start. By leveraging leading AI tooling (i.e., Claude Code), we are able to put a usable, clickable site in front of them and collect real feedback in hours or days, not weeks or months. 

That tight loop means we validate assumptions with customers in high fidelity (i.e., working prototype instead of requirements document or Figma designs). And, by the time a concept graduates to development, it has already survived contact with the people who'll actually use it, ensuring less re-works and features that resonate immediately.

‍

Delivery: AI-first workflow

Once direction is locked with the customer, our engineering workflow is specification-driven and AI-assisted from end to end:

1. Create specification & acceptance criteria. 

• Build Specification: Every effort starts with a detailed build spec, the shared source of truth for what we're making and why. The POD has been defining this specification throughout the discovery process, leveraging proprietary Claude Skills to synthesize and refine notes and concepts iteratively. 
• Acceptance Criteria: Alongside the Build Specification, the POD defines the acceptance tests and other performance criteria that must be validated before the feature can be released back to the context owner for review

2. AI agents execute the plan. 

• Our AI Agents handle the bulk of implementation against the spec, doing the work that used to consume most of an engineer's day. It utilizes the plan, our design system, our proprietary skills, and our platform capabilities to deliver on the acceptance criteria. 

3. FDE audits and reviews.

• The FDE reviews the output critically. Validating correctness, design, and edge cases rather than typing every line. If there is an issue or discrepancy, we can update our plan or add an additional plan and re-run the cycle.

4. Automated testing.

• Comprehensive automated tests based on the acceptance tests are created and run as a constant safety net, catching regressions or issues when we make meaningful changes to the code. 

5. Context Owner review.

• Once the build fully passes step 1 to 4, the Context Owner validates that what was built actually solves the original problem for the customer, closing the loop back to where we started. 

6. Security throughout.

• We've built security directly into how we use AI and LLM’s — with controls that prevent PII leakage and keep sensitive data protected at every step.

The result is a process where AI accelerates the work, humans own the judgment, and the customer's needs stay in view the whole way through.

‍

Our Results

Our new AI-native build process is under continuous refinement. So far, we have seen a range of outcomes depending on the nature and complexity of the deliverable requirement.

That said, on average, we have seen the following improvements to date: 

• ~45% faster implementations: we get from idea to working product in half the time, reducing implementation costs for customers.
• No increase in regression bugs: speed has come without sacrificing reliability or quality. We believe this is due to the ability to quickly automate tests with AI.

From our perspective, by letting AI do what it does best and freeing our team to focus on context and judgment, we've raised our velocity and maintained our standards at the same time.

This is just the start. We keep refining how PODs, FDEs, Platform Engineers, and AI work together and we'll keep measuring it against our north star: the value we deliver to our customers.