How to Implement AI in an Organization Without Losing People Along the Way.

Only 1% of organizations worldwide are defined as AI ready. Not because the technology is immature, but because adopting it inside an organization is an entirely different story from your personal adoption on your phone. This guide is the written summary of meeting 30 in the AI Master club, with Naama Maroz, artificial intelligence lead in the Innovation and Technology Administration at the Ministry of Education. The goal is to give you a complete mental framework on how AI enters an organization, where it gets stuck, and how you, in any role you hold, can lead it.

AI does not get stuck on technology. Models improve day by day. It gets stuck on adoption: fear of making mistakes, and a mismatch between what the organization needs and what AI actually gives it. AI adoption is not a one-time event, but three processes happening simultaneously: a change in workflows, acquiring new tools and skills, and a cultural mindset shift, which is the most complex and usually the most neglected.

The Reality on the Ground: The Gap Between Employees and the Organization

The opposite direction also exists, and in an extreme way: Naama met a company with only 6 employees that operates 47 AI agents, which generate most of the company value by themselves. Flat hierarchy, few people, many agents. This is not the majority yet, but it shows what the other end of the scale looks like.

Four Organizational Strategies for AI Implementation

The Four Strategies

• Button for soup: Start with one tool, and gradually thicken it, If only we had Claude Code too... Advantage: Natural and non-scary growth. Risk: You might stay almost there without a defined toolset.
• Leading horse: Focus on one tool or a set of tools (for example Copilot or Google) and specialize in it deeply. Advantage: Specializing in one tool is worth more than a superficial acquaintance with a thousand tools. Risk: Less flexibility if the tool does not fit every need.
• Jungle: No limitations, everyone uses what they want, and the organization learns from what works in the field. Advantage: Amazing employee experience, real bottom-up innovation. Risk: Organizational security loosens completely.
• Flooding: Everyone gets approval for all tools at once. Advantage: Maximum access with no barriers. Risk: In the end, no one knows what to focus on.

A side warning: Beware of token maxing, measuring productivity by the amount of tokens consumed. This leads to an unclear billing at the end of the month, and does not necessarily indicate real ROI. Silicon Valley has already taken a step back from this too.

How to Choose the Right Strategy

The Three Levels Where You Can Lead

1. Organizational Level (Macro), Centralized Approach

2. Departmental Level

• Appoint an owner who connects headquarters to the field, raises needs and rolls out capabilities
• Map core processes: What AI can do alone, what requires human in the loop, and what only humans can do
• Plan pilots that can be scaled up, not one-off pilots
• Run role-based training, not how to write a prompt, but how AI specifically upgrades the role
• Identify and nurture champions, change agents within the department

Failure points: Implementing AI on unmapped processes is shooting in the dark, wasting time and resources. Use cases that cannot be expanded later. And too superficial measurement, how many users use the tools is important, but not enough. You must prove real value to the organization.

3. Individual Level (The Person), Including Independents and Freelancers

• Literacy and mastery: Remember that not everyone is on the same page, and even basic training is still relevant for some people
• Positive FOMO: Create a feeling that there is an organizational statement and that it is worth being part of it
• One-on-one consulting: Whoever knows provides individual guidance
• Communities: A space for questions and learning that creates a sense of belonging
• Builders, not bottlenecks: It is better for people to know how to build processes themselves, rather than having everything go through IT

AI Master club opens the door for you: A live meeting every week, recordings, and a community. 47 ILS per month, cancel anytime.

Failure points: Flooding employees with tools and training without a connection to the desired result, and creating a technological bottleneck at IT/IS.

Change Management: It is About People, Not Tools

Five Stages for Creating Real Change

• Choose a goal: Define what success will look like, even at the personal level
• Choose a target audience: Who exactly you want to influence
• Choose a direction: Top-down (from management) or bottom-up (from the field)
• Define minimum requirements: What must happen for you to say we succeeded
• Measure: Not just usage, but also outputs, workflows, and milestones, and decide in advance how exactly you will measure (survey, logins, actual use)

No one will stop you

Those who succeed take small projects and upgrade them each time. Those who have the hardest time build huge projects all at once, and feel there is a problem with the technology. There is not. This is a learning curve, and that is okay.

Frequently Asked Questions About AI Implementation in Organizations

No. The problem is almost never the technology, which improves very quickly. The problem is a gap between what the organization needs and what it actually implements: fear of mistakes among employees, and a lack of orderly strategy by the organization.

No. Strategy is custom tailoring. It depends on your position in the organization, the level of attention the topic receives with you, and how many resources the organization is willing to give. You check the situation, and then choose between button for soup, leading horse, jungle, or flooding.

Yes, and that is precisely the main point in the meeting. There are three leadership levels: organizational (strategy, steering committee, toolset), departmental (process mapping, pilots, role-based training), and individual (literacy, consulting, community, independent building). Every employee can lead at the level that suits their position.

AI Sandbox is a sandbox with defined guardrails, a space where you can run safe pilots without waiting for full approval of every new tool in the organization. This is a solution to the gap between what headquarters plans and what actually happens on the ground.

Measuring productivity by the amount of tokens consumed, which turned out to be a misleading metric: it leads to unclear charges and does not necessarily indicate real ROI. Silicon Valley has already taken a step back from this, so it is better to measure actual outputs rather than usage amount.

Only 1% of organizations are defined as AI ready (McKinsey, World Economic Forum), but 79% of employees adopting AI are Gen Z, and 48% of employees are afraid to admit they use AI at work. Employees are already there, under the radar. The organization is not yet.

The Guest of the Meeting

Naama Maroz leads strategy, implementation, and change management processes in the field of artificial intelligence. She is responsible for leading systemic processes on a nationwide scale, from the stage of needs mapping and strategy formulation to solution development, team training, implementation, and impact measurement, and she manages the AI in STEM project in the Innovation and Technology Administration at the Ministry of Education.

Artificial Intelligence Lead, Innovation and Technology Administration, Ministry of Education

Implementing AI in an organization is not a technological project, it is a people project. Whoever understands this gets ahead, and whoever waits for instructions from above stays behind. Choose one level you can influence, and start.

Frequently Asked Questions

Why are most organizations still not ready for AI?

No. The problem is almost never the technology, which improves very quickly. The problem is a gap between what the organization needs and what it actually implements: fear of mistakes among employees, and a lack of orderly strategy by the organization.

Is there one strategy recommended for every organization?

No. Strategy is custom tailoring. It depends on your position in the organization, the level of attention the topic receives with you, and how many resources the organization is willing to give. You check the situation, and then choose between button for soup, leading horse, jungle, or flooding.

I am not a senior manager, do I even have a role in AI implementation in the organization?

Yes, and that is precisely the main point in the meeting. There are three leadership levels: organizational (strategy, steering committee, toolset), departmental (process mapping, pilots, role-based training), and individual (literacy, consulting, community, independent building). Every employee can lead at the level that suits their position.

What is an AI Sandbox and why does it solve a real problem?

AI Sandbox is a sandbox with defined guardrails, a space where you can run safe pilots without waiting for full approval of every new tool in the organization. This is a solution to the gap between what headquarters plans and what actually happens on the ground.

What is token maxing and why is it problematic?

Measuring productivity by the amount of tokens consumed, which turned out to be a misleading metric: it leads to unclear charges and does not necessarily indicate real ROI. Silicon Valley has already taken a step back from this, so it is better to measure actual outputs rather than usage amount.

How many organizations are actually ready for AI today?

Only 1% of organizations are defined as AI ready (McKinsey, World Economic Forum), but 79% of employees adopting AI are Gen Z, and 48% of employees are afraid to admit they use AI at work. Employees are already there, under the radar. The organization is not yet.