Product Architecture • December 2025

Why Jira, Notion, and Google Docs Cannot Support Agentic Engineering

For the past decade, product teams have relied on Jira for tickets, Notion for documentation, and Google Docs for PRDs. These tools enabled human collaboration — but they cannot support the next era of software development: Agentic Engineering.

Agentic Engineering requires a world where product intent is machine-readable, continuously discoverable, and executable. This article explains why traditional tools break in this new paradigm and why teams must transition to structured, MCP-delivered specifications.

The Core Problem: Traditional Tools Produce “Dead Text”

Jira, Notion, and Google Docs were designed for human readers, not AI agents. They store product intent as unstructured or semi-structured prose — which means:

Agentic Engineering changes the requirements of a PRD:

Agentic requirement

Product intent must be executable data — not text hidden inside documents.

Legacy tools cannot enforce structure, cannot guarantee version consistency, and cannot deliver the exact requirement an AI agent needs at the moment of code generation.

Why Google Docs Fails Agentic Engineering

1. No canonical structure

In Google Docs:

AI result:

LLMs hallucinate missing constraints because Google Docs provides no schema.

2. No versioned machine-readable fields

AI agents require strict semantics:

Google Docs offers none of these. Everything is a string blob.

3. Zero integration with development environments

AI agents in the IDE need to fetch:

Google Docs requires a human to read → summarize → paste → interpret.

Why Jira Fails Agentic Engineering

Jira is closer to the engineering workflow, but it still breaks under Agentic requirements.

1. Issue descriptions are unstructured

Consider a typical Jira story: 


As a user, I want to upload a profile picture easily.

Acceptance Criteria:
- Allows JPG and PNG
- File size limit
- Should be intuitive

This is unreadable to an AI agent because:

2. Jira cannot represent executable predicates

Tests require logic, such as: 

{
  "predicate": "file.size <= 5000000"
}

But Jira only supports freeform text or custom fields that cannot enforce logic or semantics.

3. Jira cannot push product intent into IDEs

Agentic Engineering requires:

Jira does none of these.

Why Notion Fails Agentic Engineering

Notion adds structure, but not semantics.

1. Inline databases ≠ machine-readable logic

A Notion table can store:

But it cannot store:

2. No interoperability with MCP

Notion cannot expose PRDs via endpoints that:

Notion is a knowledge base — not a product intent API.

The 6 Technical Requirements of Agentic Engineering

To support AI agents, a PRD system must satisfy six non-negotiable requirements.

RequirementWhy It MattersLegacy Tools Fail?
1. Canonical schema AI needs predictable structure ❌ Docs / Jira / Notion
2. Machine-readable acceptance criteria AI needs predicates, not prose ❌ All
3. Example-driven data AI uses examples to generate tests ❌ All
4. IDE integration via MCP Agents must pull requirements automatically ❌ All
5. Version-based traceability AI must know which PRD version applies ❌ All
6. Executable verification Tests must derive from PRD ❌ All

The New Model: Agentic Engineering Requires Executable Specifications

Agentic Engineering rejects “documents” entirely. It requires Executable Specifications, where:

This is the core idea behind:

Example: Traditional vs. Agentic Story

❌ Traditional (Jira, Notion, Docs)


User should see “PAID” badge on priced items.
Max price is high.
Should feel intuitive.

✅ Agentic PRD

{
  "id": "S-100",
  "title": "Show paid badge",
  "acceptance": [
    {"type":"predicate","expr":"price > 0 implies badges.includes('paid')"}
  ],
  "examples": [
    {"price":199,"badges":["paid"]},
    {"price":0,"badges":[]}
  ]
}

AI can build tests, generate code, and check behavior — automatically.

Why MCP Is the Missing Layer

MCP (Model Context Protocol) provides:

Example IDE Request


GET /mcp/prd/prod-2025-payment-v1/stories?file=src/components/search.ts
Authorization: Bearer <MCP-TOKEN>

The Business Impact: 35–60% Faster Delivery

Agentic Engineering isn't about aesthetics — it's about economics.

Impact areas:

Jira, Notion, and Google Docs were never designed for this world.

Conclusion: You Cannot Build Agentic Engineering on Legacy Tools

Traditional PM and documentation tools are optimized for human consumption. Agentic Engineering is optimized for AI-first execution.

Teams that continue relying on Jira / Notion / Google Docs will face:

Teams that adopt Agentic PRDs + MCP will:

Bottom Line

Agentic Engineering requires structured, machine-readable product intent delivered directly into the IDE via MCP. Legacy tools simply cannot meet these requirements.

Read Next