AI Strategy Review

AI agents coordinate through the same infrastructure as humans. That is the product.

3/5

Maturity

LOW

Displacement risk

AI agent teams active

Algorithms in platform

The AI Thesis

Stackmates is not just AI-assisted — AI agents ARE first-class citizens. The same domain model, planning DB, and commissioning logic that humans use is what agents use. This means the infrastructure gets smarter as agent capability improves. Displacement risk is LOW because we are the infrastructure being displaced into.

Function	Tool	Task	Hours/Mo	Quality
Agent coordination	Claude + A2A protocols	AI agents operate through same logic as human workflows	ongoing	HIGH
Code generation	Claude Code, GitHub Copilot	Feature scaffolding, test generation, migration authoring	20-30	HIGH
Planning DB	Claude + Convex	Task creation, plan advancement, commissioning tracking	5-10	HIGH
Content pipeline	Claude	Draft articles, meta posts, berley content for ventures	8-12	MEDIUM
Testing + QA	Claude + Playwright	E2E test authoring, commissioning evidence collection	4-8	MEDIUM

Without AI code generation, a solo architect cannot build and maintain 168K lines across 7 ventures.

Prompts

Tools

Data Prep

Workflows

Governance

Diagnosis

Strong at tools and prompts. Governance is the gap — no formal audit trail for agent decisions. Fix this before external teams onboard.

Use Case Scoring

Use Case	Impact	Feasible	Speed	Adopt	Score
Agent coordination infrastructure	5	5	5	4	19
Automated code generation pipeline	5	4	5	5	19
AI-powered commissioning triage	4	4	4	4	16
Planning DB with AI task routing	4	3	3	4	14
Venture performance prediction	3	2	2	3	10

Code generation (current leverage)

AI writes the whole stack from scratch

2-3 yearsStackmates IS the infrastructure — generates into it

Domain model (10-layer hex)

AI generates domain models from specs

3-5 yearsBecome the spec — own the commissioning layer

Agent coordination layer

Platform providers (Anthropic, OpenAI) build native A2A

1-2 yearsIntegrate early, don't compete — be the domain on top

Displacement risk is LOW because Stackmates IS the infrastructure. The question is whether the domain model on top remains the moat.

Maturity Roadmap

Ad hoc

Individual tool experimentation

DONE

Emerging

Defined use cases, agents in worktrees

DONE

Defined

Repeatable agent workflows, documented protocols

CURRENT

Managed

Automated pipelines, quality metrics across ventures

Target q4

Optimised

Self-improving infrastructure, agents improve templates

The Differentiator

Most platforms treat AI as an add-on. Stackmates treats AI agents as first-class infrastructure users. When an agent commits code, runs a plan, or files a commissioning update — it uses the same domain model a human does. This means AI capability improvements compound directly into infrastructure value. The moat is not the code. It is the domain model that AI and humans share.

Agent runs task→Uses domain model→Advances plan→Updates ledger→Infrastructure learns

Questions

If A2A protocols become a commodity in 12 months, what remains Stackmates' unique layer?

Is the domain model defensible if an AI can generate one from a spec in 10 minutes?
Which of the 13 algorithms becomes obsolete first as AI capability improves?
At what AI maturity level does external onboarding become fully self-serve?