The builder pattern

SCRUM for AI-assisted collaboration.

Developers are using AI to write code. Operators are using AI to run their businesses. Teams are using AI to draft plans, write policies, and make decisions. Almost none of it is structured, reviewed, or visible to the rest of the team. The builder pattern is how p6k closes that gap — a universal, structured process for how people and AI collaborate to produce pluggable, reviewable output.

The promise

Something fundamental is happening in how work gets done. People are using Claude, ChatGPT, and Copilot to do the actual job — and the output of that work lives in chat logs nobody can see, track, approve, or improve. Collaboration is missing. Visibility is missing. Review and approval are missing.

This is not just a software development problem. It is a structured collaboration problem. Whenever multiple people work with AI to produce something — code, configuration, documentation, or compliance forms — the same gaps appear. The builder pattern is the answer.

The flow

Six steps. One universal loop.

Every builder runs the same shape. Only the output changes.

  1. 01

    Interview

    AI-driven, multi-person knowledge capture. The system asks questions. The people who run the work answer them. The conversation exposes the business process, requirements, or design — not in a documentation sprint, but at the point of work.

  2. 02

    Structure

    The captured knowledge is organized into a model — playbooks, processes, requirements, data, whatever the domain calls for. The raw material of the interview becomes something a machine can reason about.

  3. 03

    Generate

    The model is rendered into artifacts. Configurations today. Code, documents, compliance forms tomorrow. The output type is pluggable — defined by the process itself, not hardcoded into the platform.

  4. 04

    Review

    Humans verify, approve, or request changes. The right people sign off before anything goes live. AI-assisted work produces auditable artifacts, not chat logs nobody can govern.

  5. 05

    Publish

    Artifacts go live. Configs get deployed, code gets merged, documents get published. Work moves from draft to canonical with a single, shared process — no matter what the output is.

  6. 06

    Evolve

    The builder is a continuous loop, not a one-time event. As the business grows, requirements change, or the team learns more, the whole flow runs again. The system gets smarter every time the business uses it.

Before and after

Before p6k

Your team uses ChatGPT to draft policies, Claude to write code, and Copilot to suggest completions. Nobody can see what anyone else captured. Decisions live in chat logs. Nothing gets reviewed. The output of AI-assisted work is a pile of artifacts with no process around them.

After p6k

Every piece of AI-assisted work runs through the same structured loop. Interviews expose knowledge. Structure organizes it. Generation produces artifacts. Review approves them. Publishing makes them canonical. Evolution keeps them current. The AI does the work. The builder makes it a system.

Same flow. Different output.

One pattern. Every kind of structured work.

The builder process is universal. The output is pluggable. That is what makes it a platform, not a feature.

Output · p6k app configurations

A specialty insurance brokerage

An owner-operator is interviewed about how his business runs. His team contributes answers and fills in gaps. Out the other side come tables, workflows, pages, and business rules — software tailored to his company without a consultant in the room.

Output · ITIL-flavored configurations

A managed services provider

Same builder flow, different context. The interview is shaped by ITIL best practices. The output is still p6k configurations, but reflecting the MSP’s specific operational model.

Output · Source code

The p6k team building p6k

The same interview, structure, generate, review loop — but the artifact is source code. Architecture decisions and implementation plans are captured collaboratively, reviewed, and turned into a working codebase.

Output · SOPs, PDFs, compliance forms

A biotech operations team

Stakeholders are interviewed. Process knowledge is structured. Documents are generated. Reviewers approve before publication. The same flow that produces code can produce regulatory submissions.

The pitch, in one paragraph

The builder pattern is what SCRUM was for agile development — the framework the industry needed but did not yet have a name for. A structured process with defined roles, versioned artifacts, review ceremonies, and continuous improvement. Not ad-hoc prompting, but a shared vocabulary for teams working through AI. This is not a feature of p6k. It might be p6k.

Proof points

  • Universal process. Pluggable output. Configs, code, documents, compliance — same flow, different artifact.
  • Multi-person by design. Contributors, reviewers, and approvers each have clear roles and visibility.
  • Governance built in. Every artifact is versioned, reviewed, and traceable — not buried in a chat log.
  • Continuous improvement. The evolve step means the system gets better every time the business runs it.

The bet

The operating system for AI-native teamwork.

Most AI companies are building features. p6k is building the platform whose core primitive is structured AI-assisted collaboration. If the builder pattern is right, it changes what a business operating system can be.