---
title: "Which stage of AI are you really at?"
description: "5 source-backed AI claims from Which stage of AI are you really at?, with key statistics, context, and the original AI Adopters Club source."
url: "https://kbanc.com/claims-library/which-stage-of-ai-are-you-really-at"
source: "https://aiadopters.club/p/which-stage-of-ai-are-you-really"
date: "2026-07-20"
topics: ["strategy", "implementation", "measurement"]
generated: "2026-09-19"
---

# Which stage of AI are you really at?

By Kamil Banc | July 20, 2026

## Claims

1. **Agent Count Is Vanity**: Agent count is a vanity metric that reveals nothing about actual AI adoption maturity.
2. **Maturity Varies By Workflow**: AI maturity operates per-workflow, allowing a company to be advanced in engineering but blocked in finance.
3. **Productivity Feels Up, Quality Down**: In a 2026 study, 84% of software engineers felt more productive despite work quality declining.
4. **Operating-Model Change, Not Software**: Most AI adoption failures result from treating implementation as software rollout rather than operating-model change.
5. **Self-Assessment Prompt Scores Readiness**: A self-assessment prompt scores organizations across access, governance, verification, cost, skills, and risk factors.

## Evidence

### Quote
> "A hundred agents nobody's checking isn't maturity. It's a bigger invoice and a slower incident." - Kamil Banc

### Key Statistics
- **84%**: Percentage of software engineers in a 2026 study who felt more productive with AI despite the work itself becoming harder and less flowing
- **Five stages**: Boris Cherny's original AI adoption ladder categorizes maturity by agent count: zero, one, ten, a hundred, and a thousand-plus
- **30 to 90 days**: Timeframe the self-assessment prompt uses to generate an action plan with measurable exit criteria for clearing the identified gate

## Context
The article reframes AI maturity assessment around five gates—access, trust, capacity, governance, and economics—rather than the popular agent-count ladder popularized by Anthropic's Boris Cherny. Kamil Banc argues that most organizations fail not due to weak AI models but because they treat adoption as a software rollout instead of an operating-model change requiring redesigned workflows, ownership, and measurement. Practitioners are encouraged to use a structured self-assessment prompt that evaluates their setup across seven dimensions and outputs a specific 30-90 day plan with measurable exit criteria. A supporting comment from a media company practitioner corroborates the framework, noting that trust and capacity gates are where most real-world stalls occur, particularly around audit trails and review capacity for specialized data like rights and licensing.

## Source
- Original: [Which stage of AI are you really at?](https://aiadopters.club/p/which-stage-of-ai-are-you-really)
- Cite: kbanc.com/claims-library/which-stage-of-ai-are-you-really-at