Culture still eats tooling

The most expensive AI failures often begin with leadership behavior.

A founder says the company must become AI-first, then praises the flashiest demo instead of the team that improved a real outcome. A CTO encourages everyone to try AI, then punishes the first group that raises a security concern. A COO asks every department to automate, but never names which workflows matter or who owns the results.

The organization hears the real message: move fast, hide risk, show progress, and do not ask too many questions about outcomes.

That is how companies create speed theater — AI activity without operating improvement. Demos multiply. Dashboards glow. Agents appear in workflows. Leaders celebrate adoption before they understand whether the work got better.

The problem is not only tool cost. The bigger cost is management debt. Teams learn that visible AI activity matters more than accountable learning, governance is something to route around, and leadership wants momentum more than truth.

Culture still eats tooling. It always did.

Culture mattered before AI

Before AI, culture already determined whether technology created value. A company could buy the best project-management system and still avoid hard prioritization. It could adopt agile ceremonies and still punish honest estimates. It could install a modern data platform and still argue from opinion.

The written system says one thing. The lived system says another.

People are not trained by values posters. They are trained by what leaders reward, tolerate, and ignore. Culture shows up in who gets attention, what gets measured, what gets excused, how conflict is handled, and whether truth can travel upward without punishment.

In the original book, I wrote about great culture, care, consideration, and prioritizing life. I still believe that. It is not soft. It is operating infrastructure.

People do better work when they understand the mission, trust the rules, and believe leadership will not punish reality. That mattered when teams were shipping software. It matters more when AI can accelerate every pattern inside the company.

AI amplifies the culture you already have

AI does not create a new culture from scratch. It reveals the one you have.

In a culture of clarity, AI adoption becomes focused. Teams ask what outcome they are improving, connect tools to workflows, define review standards, name owners, and treat failed pilots as evidence.

In a culture of fear, AI goes underground. People use tools quietly because asking feels slow or risky. Shadow automation spreads.

In a culture of theater, AI becomes a demo factory. Teams optimize for screenshots. Leaders mistake activity for progress. Workflow redesign, data cleanup, decision rights, quality standards, and exception handling get postponed because they look less exciting than the demo.

In a culture of accountable learning, AI becomes a management discipline: here is what improved, here is what did not, here is what we learned, here is the risk we found.

That is the fork: shadow automation or governed leverage.

Leaders design the signals

A modern AI operating system is made of tools, workflows, agents, policies, scorecards — and signals.

Every leader sends signals all day. What you ask about, ignore, praise, fund, and forgive teaches people how the system really works.

Ask only “How many AI use cases launched?” and teams optimize for count. Ask “Which business outcome improved, and what evidence do we have?” and teams start connecting AI work to outcomes.

The principle is:

AI adoption follows leadership signals before it follows tooling strategy.

Leaders need to make desired behavior visible in daily management:

• Direction before speed.
• Outcomes over tasks.
• Workflow before agent.
• Human accountability for AI-enabled work.
• Learning over performance theater.
• Sustainable execution over heroic exhaustion.
• Candor about risk before risk becomes an incident.

These become culture only when they change what people do on a normal Tuesday.

Care is part of the system

Care is not avoiding change or lowering standards. Care is designing work so people can meet high standards without burning judgment as fuel.

AI adoption can create real anxiety. People wonder whether their work is valued, whether output expectations will rise faster than review capacity, and whether mistakes made with AI will be treated as learning or personal failure.

Leaders should address that directly. If an agent creates one hundred drafts a week, someone reviews them. If coding agents help engineers ship more code, architecture and review standards must keep up, and so must the reviewers. If automation changes a workflow, the people inside that workflow need a voice in the redesign.

Ignoring this creates hidden exhaustion. Teams may appear to adopt AI while quietly absorbing more review work, coordination work, and ambiguity.

Sustainable execution is an operating advantage. A tired team may comply with an AI mandate. A trusted team will help build the system that makes AI useful.

Culture signals checklist

Before scaling AI across a team or workflow, inspect the signals:

1. Reward signals

• What AI behavior has been praised publicly in the last thirty days?
• Did we praise output volume or measurable outcome improvement?
• Did we recognize teams that surfaced risk early?
• Which official KPI is this AI change supposed to improve, who owns it, and where is it reviewed?

2. Permission signals

• Do people know what AI use is allowed without approval, including which tools, connectors, and data an agent may use?
• What requires review?
• What is prohibited?
• Are teams punished for asking governance questions early?

3. Accountability signals

• Does every important AI-enabled workflow have a human owner?
• When an agent makes a mistake, who corrects it and captures the learning?
• Are informal incentives pulling teams away from the official scorecard?

4. Learning signals

• Where do teams share what AI pilots are teaching them?
• Are failures documented in a useful way, and do they become test cases for the next version?
• Do we have a cadence for continue, change, scale, or stop decisions?

5. Care signals

• Where is AI increasing review burden or cognitive load?
• Are people trained for the new workflow?
• Do employees have a safe way to raise concerns about quality, ethics, security, or workload?

6. Leadership consistency signals

• Do executives follow the same standards they ask teams to use?
• Do leaders ask for evidence before scaling?
• Are tradeoffs explicit, or hidden inside urgency?

You can run this in less than an hour. The value is exposing the gap between the culture you declare and the culture your AI adoption is living inside.

A short example

A support team pilots an AI assistant that drafts replies. The demo is strong. Drafts appear in seconds. Leadership praises speed.

Then edge cases appear: contractual commitments, product defects, outdated documentation, confident but wrong replies. Support managers review more carefully, but review work is not in the capacity plan. The team reports time saved because that is what leaders asked for, while privately spending more time correcting drafts.

A speed-theater culture keeps pushing and tunes the prompt. An accountable-learning culture pauses to inspect ticket types, knowledge-base ownership, escalation rules, review capacity, and success metrics beyond first response time. It turns the replies that went wrong into a small eval set and runs every change against it.

The tool did not decide which culture the company had. Leadership did.

Chapter 3 operating check

Ask:

1. What AI behaviors are we rewarding: speed theater, tool usage, accountable learning, measurable outcomes, or risk reduction?
2. Where are people likely using AI quietly because the official path is unclear, slow, or unsafe?
3. Which pilots show activity but weak business or customer impact?
4. Where have we increased AI output without review capacity or quality standards?
5. What governance questions do teams hesitate to raise?
6. Which leader behaviors conflict with stated values?
7. Where is AI increasing cognitive load, rework, or coordination work?
8. Which workflow needs clearer ownership before scaling?
9. What signal can leadership change this week?

The culture you have will become the AI operating system you get. Leaders cannot delegate that to a platform.