Neomanex Logo
ConvOps Manifesto

Conversations are where work happens. It's time to give them a process.

ConvOps (Conversational Operations) is the practice of driving work through conversations that follow a defined workflow.

What is ConvOps?

Three roles, clearly separated.

The Orchestrator

The AI agent that drives the conversation — the one the human talks to. It reads the current step, executes or delegates the work, and advances.

The Conversation

Where the work happens — the execution environment. Not a pipeline. Not a canvas. A conversation.

The Workflow

Defines how the work should be done — steps with instructions, gates where humans are required.

"Chatbots structure conversations to answer questions. ConvOps structures conversations to get work done — step by step, through a defined process."

How ConvOps Is Different

Chatbot frameworks
Structures: Conversations that answer questions
How: Dialog trees, intents
Agent frameworks
Structures: Code that orchestrates agents
How: DAGs, state machines
Workflow tools
Structures: Pipelines that automate steps
How: Visual canvas, triggers
Agentic workflows
Structures: Pipelines where agents execute steps
How: Agent → output → next agent
ConvOps
Structures: Conversations that follow a process
How: Workflow steps embedded in the conversation

The Principles

Nine principles that define how ConvOps works.

1

The orchestrator drives, the workflow guides

The orchestrator — the AI agent the human talks to — drives the conversation. It reads the current step, executes or delegates the work, and advances when done. The human directs. The orchestrator executes. The workflow surfaces the current step’s instructions inside the conversation.

2

Conversations without process are chaos

Unstructured conversations produce inconsistent results. When a conversation follows a defined workflow — step by step, with clear instructions at each stage — the output is consistent, repeatable, and auditable. Every time.

3

The step surfaces inside the conversation

Each step appears inside the conversation as a briefing — what to do, what standards to follow, what to produce. The human or agent doesn’t leave the conversation to check a dashboard. The step comes to them, right where they’re working.

4

Gates define where humans are required

A gate is a point in the workflow where the process defines whether a human needs to be involved. A gate might require approval, review, handoff, or sign-off. When a gate exists, the conversation pauses until the human acts.

5

Any agent can connect

ConvOps is not tied to one AI provider, one model, or one tool. Any AI client — coding agents, chat assistants, custom agents, or humans themselves — can connect, read the current step, and execute. The workflow is the contract.

6

Conversations are resumable

Because every step records what was done and what comes next, anyone — human or agent — can pick up where the last one left off. Context is never lost. Conversations never start from zero.

7

The workflow is the audit trail

Every step records who did the work, what was produced, and what decisions were made. The conversation that executed the work IS the record of the work. Visible and auditable by design.

8

Process knowledge is encoded, not tribal

Company standards, best practices, regulatory requirements — these live in the workflow step instructions, not in someone’s head. People leave. Process stays.

9

The workflow replaces the prompt dump

Today, entire processes are crammed into system prompts — pages of instructions with no structure. ConvOps replaces this with a workflow where each step delivers only the instructions needed right now.

The Main Thread

Every ConvOps workflow runs inside a main thread — the orchestrating conversation that drives the work. The main thread surfaces the current step, tracks progress, and keeps the overall narrative clean and readable.

In environments that support sub-agents, the main thread delegates execution to workers and stays high-level. In simpler environments — like a chat with Claude or ChatGPT — the main thread does everything itself.

The workflow doesn't change. The steps don't change. The gates don't change. The main thread just adapts to what the environment can do.

ConvOps in Practice

A tech lead needs a feature built. With ConvOps, they define a workflow.

Step 1

Analysis

Brainstorm approaches. Consider the existing architecture. Write the analysis.

Step 2

Create Plan

Write a detailed implementation plan with phases, dependencies, and estimated effort.

Step 3

Plan Review

Review the plan for architectural issues, security concerns, and alignment with standards.

Human gate
Step 4

Fix Plan

Address the feedback from the review. Update the plan with revisions.

Step 5

Execute Plan

Implement the plan phase by phase. Follow coding standards. Commit after each phase.

Step 6

Done

Work complete, audit trail preserved.

The developer picks up the task and starts a conversation with their AI assistant. The workflow surfaces each step inside their conversation. When a gate is reached, the conversation pauses until the human acts. The process is enforced, auditable, and resumable.

When the tech lead later adds a testing step, every future conversation reflects it automatically. No retraining. No announcement. The process updated, and every conversation reflects it.

What ConvOps Replaces

Employee opens ChatGPT, asks random questions
Employee starts a conversation, workflow guides the process
Agent gets a prompt dump with the entire process
Current step surfaces inside the conversation with clear instructions
Every person does it differently
Every conversation follows the same process
No audit trail of AI-assisted work
Workflow steps record who did what and when
Manager has no visibility
Manager sees workflow progress and gate approvals
Work can’t be resumed by someone else
Any person or agent picks up at the current step
AI usage is ungoverned
Gates define where humans are required
Process knowledge lives in people’s heads
Process knowledge lives in workflow step instructions

Who ConvOps Is For

Teams adopting AI

Your people are already using AI through conversations. ConvOps gives those conversations structure, consistency, and governance — without changing how people work.

Developers building with AI agents

Your agents execute based on prompt dumps. ConvOps replaces that with a workflow — step-by-step instructions, gates where humans are required, and a conversation that serves as both the execution environment and the audit trail.

Leaders responsible for AI governance

You need to know what your AI agents and employees are doing, whether they’re following process, and where human review happened. ConvOps makes AI-assisted conversations visible, auditable, and governed.

ConvOps Is Not

Not a chatbot framework. Chatbots structure conversations to answer questions. ConvOps structures conversations to follow a process.

Not another agentic workflow. Agentic workflows are pipelines where agents execute steps. ConvOps embeds the workflow inside the conversation — the orchestrator drives through a living conversation with full context.

Not a workflow automation tool. There’s no visual canvas, no drag-and-drop pipeline. The conversation is the execution environment.

Not an agent framework. ConvOps doesn’t build agents. It gives any agent — or human — a process to follow inside their conversation.

Not a project management tool with AI bolted on. ConvOps is conversation-first — built for how work actually happens now.

Ready to AI Governed?

Read the full manifesto on GitHub. Open source. Contributions welcome.

Conversation-First
Human Gates Built In
Auditable by Design