#2: The Unsexy Truth of AI Adoption

Turing Post (Ksenia Se) · email · 2026-04-04 source ↗ Reader 🖥️

Enterprise AI Adoption & Readiness Organizational Legibility

AI adoption in organizations is not a technology selection problem but an organizational legibility problem: companies must make their tacit knowledge explicit, verifiable, and structurally trustworthy before AI can compound, and the hardest transition — making the organization legible to itself — is consistently underestimated and skipped by companies that rush toward autonomous AI capabilities.

practical · mixed — Practical: argues organizations must build through specific maturity levels in a defined order. Expository in form (framework + case studies), but the framework is prescriptive. The case studies are the strongest part — two detailed examples of L1→L2 work. The rest of the framework (L2-L5) is described but not fully elaborated in the accessible text.

Structure

Part	Role
Reality check: SF hype vs. actual company landscape	Context/motivation — establishes that most companies are far from the AI-enabled future being discussed; creates credibility by acknowledging the gap between narrative and reality
Framework introduction: AI maturity as a stack of dependencies, not a timeline	Core argument — reframes the question from 'what stage are we in?' to 'where are the structural gaps blocking us?'
What most companies actually want (skip to agents/autonomy)	Problem diagnosis — shows the misalignment between what companies want and what the foundation requires
Why the middle cannot be skipped: AI maturity is cumulative	Core structural argument — each level reveals something previously invisible, and you can't fake it
L1→L2 case study: bookkeeping company (invoice data)	Primary evidence — shows what making tacit knowledge explicit actually involves in practice (6 weeks before any AI); the 'light pushing out darkness' observation is the key insight
L1→L2 case study: construction firm (cost code mappings, project manager games)	Secondary evidence — shows the political dimension: knowledge opacity enables cover for organizational dysfunction (moving money between buckets to manage client expectations)
L2→L5 framework (paywalled)	Completes the stack: L2 (trust your data), L3 (act on signals), L4 (self-improve) — only accessible as framework-level names

Key terms

stack of dependencies Enterprise AI Adoption & Readiness
AI maturity is not a linear sequence of optional stages but a set of capabilities where each layer must be stable before the next can be built. You cannot skip L2 to get to L4 — the L4 capability will collapse. This is the core metaphor that distinguishes the article's framework from standard maturity models.
legible (to a machine) Organizational Legibility
An organization is legible when it can describe its own work in a form that a machine can act on: explicit rules, documented exceptions, named conventions, defined approval chains — the actual processes, not the policy document version. Distinct from 'visible to management': legibility requires precision machines can work with.
pilot purgatory Enterprise AI Adoption & Readiness
Organizations that keep starting AI pilots which work in isolation but never compound or connect. Caused by skipping L2: each pilot requires the knowledge that L2 makes explicit, so they remain brittle and personal — when the employee leaves, the workflow leaves with them.
tacit knowledge Organizational Legibility
Organizational knowledge that lives in people's heads, in informal practice, or in relationships — the actual rules for processing an invoice, the real approval chain (not the org chart version). Not captured anywhere, not describable in a form machines can act on. The primary blocker at L0-L1.
bus factor Organizational Legibility
Used directly: the risk when critical knowledge lives in one person. Making work legible reduces bus factor. The construction firm case (224 commits of logic in one developer's head) is the exemplar.

Propositions

explicitcorroboratedMost companies are far from autonomous AI operation — still at L0/L1: ChatGPT for writing, Copilot for code, meeting summaries, vague pressure to 'do something with AI'conf: high
supports ←Everyone wants to be AI-pilled. Most Companies Are Still Level 1 · p1 — Both sources independently converge on the same empirical claim — that despite loud AI rhetoric, most real companies remain at the lowest maturity levels (L0/L1) — using parallel but independently constructed leveling frameworks.
explicitcorroboratedAI maturity is a stack of dependencies — you cannot skip levels without collapseconf: high
supports ←The Org Age of AI: A Collection of Enterprise AI Adoption Guides · p6 — This digest's own sequencing of the eight series items is an instance of the 'you cannot skip levels without collapse' dependency structure that the fuller maturity-framework extraction states directly.
explicitThe L1→L2 transition — making the organization legible to itself — is the hardest transition in the entire frameworkconf: high
explicitcorroboratedMost companies want to skip from L1 directly to autonomous agents — and this is why pilots failconf: high
supports ←Adrien Taravant's Post · p1 — The 17%-of-149 statistic is a concrete data point for the broader thesis that stated AI ambition (everyone 'wanting/planning' one) vastly outpaces working deployment — here in the specific sub-domain of org-wide AI memory systems.
explicitMaking work legible to machines also makes it better for humans — onboarding faster, bus factor lower, organization more resilientconf: high
explicitSome institutional knowledge is deliberately kept tacit to protect indispensability and conceal organizational dysfunctionconf: high
explicitMaking work legible is a form of vulnerability for humans — it makes work inspectable and therefore contestableconf: high
explicitGovernance almost always trails deployment — different departments will be at different maturity levels simultaneouslyconf: high
explicitdevelopedAI is not an 'AI strategy' or 'tool selection' problem at L1→L2 — the blocker is that the company cannot describe its own workflowsconf: high
extends ←Everyone wants to be AI-pilled. Most Companies Are Still Level 1 · p3 — The other source identifies 'the company cannot describe its own workflows' as the real L1→L2 blocker; this thread's 'what can AI see?' question operationalizes exactly that same legibility blocker as the first of its four diagnostic axes, independently.
explicitLegibility work forces clarity that has secondary effects: reducing games played in the dark (supplier fees, budget manipulation)conf: high
implicitThe resistance to L2 is not just inertia — it is rational self-protection by knowledge holders who face real incentive problemsconf: high
explicitAI compresses the distance between large enterprises and small teamsconf: low

Arguments

The L1→L2 transition requires organizational self-description, not better AI tools strength: high

Explicit premises:

Companies at L1 look more advanced than they are — AI use is personal, brittle, undocumented
When the employee leaves, the AI workflow leaves with them
AI tools cannot absorb the ambiguity that humans have been absorbing implicitly

Implicit premises:

AI requires deterministic, machine-actionable inputs — it cannot infer unstated rules the way human workers learn them over time through socialization
The problem is not tool capability but input quality; even perfect AI tools fail on undocumented processes

Inferences:

Organizations that believe they have an AI problem may actually have a knowledge management problem
The consulting and procurement cycles that focus on 'AI strategy' and 'tool selection' systematically miss the actual blocker

Making organizational processes explicit reveals and disrupts informal power structures that depend on opacity strength: high