How to build an AI training plan that starts with leadership

In practice, a plan like this is organized by deliverables, not dates. A stage ends when it has produced what the next one needs to begin: the board delivers priorities and usage rules, managers deliver quality standards and redesigned workflows, and operations receive use cases that already have criteria and an owner. A course calendar without those deliverables is just a full schedule.

Why the order matters

The board is the one that sets direction, priorities, budget and culture. If it doesn’t use AI, it decides about AI by ear: it approves tools on the strength of a demo, accepts the pilot from the most excited area and can’t tell the team what it expects from AI-assisted work.

When operations learn first, volume rises before judgment does. The team produces faster, the manager receives more material to review without knowing how, and each area opens its own pilot. Months later, no one can say what worked.

First, AI enters leadership’s routine. Then, the way the company works.

That’s the order of Almai’s method: the board, middle management, operations, with the AI-First Executive Journey in the first stage and the AI for Middle Managers program in the second. The reason isn’t hierarchy. Each level needs what the previous one decided in order to work well.

What the board needs to deliver

The board’s stage starts with each person’s own work and ends in decisions for the company. Before moving on, it needs to have:

• Its own use: each director uses AI on recurring tasks, such as preparing the board meeting, analyzing the monthly report or reviewing a vendor proposal.
• Priorities: a short list of where AI matters to the business, with an owner and criteria for each item.
• Usage rules: what can and can’t go into the tools, and what requires human review. The generative AI use policy is the document that formalizes this.
• A common language: the board uses the same terms and knows what it’s asking for when it talks about agents, automation or pilots.

A simple test: if a director can’t show how they used AI in the past month, the stage isn’t finished.

What managers need to deliver

An AI literacy track for managers should teach less about generating and more about reviewing. The team now does in minutes the research that used to take days, and someone has to decide whether it holds up. That someone is the manager.

The stage ends when middle management has:

• Quality standards: each area defines what a good AI-assisted deliverable is, with sources checked, numbers verified and a conclusion that holds up.
• Critical review: the manager recognizes workslop, work that looks finished before it’s any good, and sends it back before it reaches the client or the board.
• Team coaching: the manager teaches from the team’s real deliverables, not from a manual.
• Redesigned workflows: at least one process per area redone with AI, with review at the right point along the way.
• A living library: prompts, criteria and lessons learned kept somewhere the team consults and updates.

What operations need to deliver

When training reaches operations, it already has direction. The priorities say where to apply it, the rules say what’s allowed, and the manager knows how to review what comes out.

Here the training is by area and about that area’s work. Finance works on reconciliation and analysis, sales on proposals and client research, legal on contract review. The stage ends with:

• Use cases by area: chosen from the board’s priorities, each with an owner.
• A shared library: the same one the managers started, now fed by the people who do the work.
• Autonomy: people adapt and create their own uses within the rules, and the manager reviews the result.

Governance runs in parallel

Governance isn’t a stage after operations. It starts along with the board and accompanies the whole plan. The usage policy is born in the first stage and revised as managers and operations bring new questions. An AI committee with senior leadership monitors the priorities, decides on pilots and adjusts the rules.

For companies operating in the European Union, AI literacy is also a legal obligation. Article 4 of the AI Act, the EU’s AI regulation, requires providers of AI systems and the organizations that deploy them to take measures to ensure a sufficient level of AI literacy among their staff, taking into account their knowledge, experience, training and the context of use. The rule has applied since February 2, 2025. If your company operates or sells there, confirm with legal how the rule applies to you.

Even outside that case, the text serves as a reference for the plan: it calls for literacy proportionate to each person’s role and context, the opposite of the same course for everyone.

Common mistakes

• Starting with operations: the team gains speed without knowing what matters, and the manager becomes a review bottleneck.
• One generic course for everyone: the board gets bored, operations don’t apply it and no one works with their own material.
• Pilots without owners: each area tests something, and no one decides what scales and what stops.
• Buying tools before deciding priorities: the license arrives before the question it was supposed to answer.
• Measuring by active licenses: access isn’t use. The guide on how to measure the results of AI training shows what to look at instead.

A plan template to copy

For each stage, write five lines: who takes part, on what real material, what it needs to deliver, who the owner is and who decides it can move on. A skeleton to start with:

1. The board: C-level executives, VPs and directors, working on the quarter’s meetings, reports and decisions. Delivers its own use, priorities and usage rules. Owner: the CEO.
2. Governance, in parallel: committee formed and usage policy published. Owner: a director named by the CEO.
3. Managers: each area’s middle management, working on the team’s recent deliverables. Delivers quality standards, redesigned workflows and the library started. Owner: the area’s director.
4. Operations: teams by area, based on the priorities set. Delivers use cases with owners, a shared library and autonomy with review. Owner: the area’s manager.
5. Plan review: what changed, what to stop and what goes into the next round. Owner: the AI committee.

The criterion for moving on applies to every line: the next stage starts when the previous one has delivered, not when the calendar says so.