An Executive AI Literacy Curriculum: A Whitepaper

Executives are asked to fund, govern, and be accountable for systems that act, and most have never been given a structured way to understand them. General AI courses teach concepts that do not convert into decisions; vendor briefings teach products. This whitepaper sets out an executive AI literacy curriculum defined by decision competence: eight modules, each ending in something the executive can do, delivered on the company's own agents, and assessed by the questions leaders ask in real reviews.

What is executive AI literacy?

The ability to make good decisions about AI systems without building them. An AI-literate executive can judge whether a pass rate is evidence, decide what an agent may do alone and on what basis, ask what an agent could do if compromised, read a cost-per-task trend, recognize theater, and communicate honestly with employees about what changes. None of this requires technical depth. All of it requires a working model of how agents work and what can go wrong. FISTA's agentic AI explained for executives piece is the first module's reading; this whitepaper sets out the whole curriculum.

What are the design principles?

1. Capability outcomes, not content lists. Each module is defined by what the executive can do afterward.
2. The company's own material. Modules use the company's agents, evidence, incidents, and processes, not hypotheticals.
3. Applied between sessions. Each module is followed by a real review in which the capability is used.
4. Assessed by observation. The questions executives ask in reviews are the test.
5. Role-specific extensions. The core is shared; functional leaders get modules for their domain.
6. A board version. Shorter, focused on oversight.

What are the eight modules?

Module	Capability outcome	Core reading	Applied exercise
1. What agents are	Distinguish agents from chatbots and automation; state the three properties that matter (they act, probabilistically, and the company stays accountable)	Agentic AI explained for executives	Classify the company's current AI systems as tools, assistants, or agents
2. How agents work	Name the five parts (model, tools, loop, context, guardrails) and ask a question about each for any agent	How AI agents work, explained for executives	Review one live agent's design against the five parts
3. Where value comes from	Identify processes with volume, rules, baselines, and contained blast radius; state the thesis beyond cost	The CEO's guide to AI and agentic AI	Score three candidate processes and name the outcome for each
4. Evaluation as evidence	Judge whether a claim of accuracy is evidence; ask for the set, the pass rate, and the date; refuse demos as proof	AI evaluation explained for executives	Run the next review opening with evaluation results; record what changed
5. Risk and controls	Name the seven risk categories and the control for each; ask what an agent could do if compromised	AI risk explained for executives	Map one live agent against the seven categories and find the unowned one
6. Autonomy and decision rights	Set autonomy per action class on consequence and evidence; state which lines policy holds; know who decides what	How much autonomy should AI agents have	Propose one autonomy change with evidence at the quarterly review
7. Economics	Read cost per task against baseline; separate build from run; recognize activity metrics	The CFO's guide to AI and agentic AI	Reconstruct one agent's cost per task and compare with its baseline
8. The human side	Redesign a role around an agent; communicate a change with candor; decide the disposition of freed capacity	How to communicate AI changes to employees	Draft the employee message for the next deployment and review it with HR

Each module runs about ninety minutes: a short briefing, work on the company's own material, and the assignment of the applied exercise. Modules are spaced two weeks apart so that each exercise happens in a real review before the next session.

Module one and two: what agents are and how they work

The foundation is a mental model precise enough to ask questions with. Executives leave module one able to explain, in their own words, that agents take actions in company systems under granted permissions, that their behavior is probabilistic and must be tested, and that accountability stays with the company. Module two adds the five-part model, so that any proposal can be interrogated part by part: what tools and permissions, what stop conditions, where context comes from, what guardrails and evidence. The glossary entry on agentic AI supports the vocabulary.

Module three: where value comes from

Executives learn to find value where agents produce it: high-volume, rule-bounded, measured processes, and to state a thesis in terms of cycle time, capacity, and services that become economical, not only cost. The applied exercise uses the AI use-case scoring framework on real candidates from the executive's own function.

Module four: evaluation as evidence

The single most valuable module. Executives learn what an evaluation set is, what a pass rate means, why a demo is not evidence, and how to ask the three questions: what is the pass rate, on how many cases, and when did it last run? The applied exercise changes the company: the next review opens with evaluation results, and executives observe how the conversation changes.

Module five: risk and controls

Executives learn the seven categories (wrong actions, security and manipulation, data exposure, drift, dependency, regulatory, reputation), the control for each, and the one question that reveals exposure: what could this agent do if fully compromised? The prompt injection explained for executives piece is the supporting reading for the security category, because it is the risk executives most need to understand and least often do.

Module six: autonomy and decision rights

Executives learn to set autonomy per action class on two inputs, consequence and evidence, within a written appetite, with lines that policy holds. They learn who decides what at each level. The applied exercise is a real proposal at the quarterly review, with evidence cited. The AI decision rights framework is the supporting artifact.

Module seven: economics

Executives learn to budget build and run separately, to read cost per task against baseline, to recognize that run cost becomes the larger line, and to distinguish outcome metrics from activity metrics. The AI funding models for executives piece supports the funding-structure discussion.

Module eight: the human side

Executives learn to redesign roles with the incumbents, communicate change with specifics and candor, decide the disposition of freed capacity after evidence, and keep failure reporting safe. The how to think about AI and headcount guide addresses the conversation most executives find hardest.

What are the role-specific extensions?

After the shared core, functional leaders take one further session on their domain, using the companion guides in this series: the CIO's guide for platform, the CISO's guide for security, the CHRO's guide for workforce, the general counsel's guide for legal, the CDO's guide for data, the CPO's guide for product, and the COO's guide for operations.

What is the board version?

A half day focused on oversight: what agents are (module one), what the board should expect from management (inventory, tiers, evidence, reporting), the questions directors should ask, and how the duty of oversight applies. Delivered on the company's real deployments, with the board director's guide to AI and agentic AI as pre-reading and the AI oversight and fiduciary duty piece as the governance reference. Directors who want the full curriculum can take it.

How is literacy assessed?

By observation in real reviews. A literate executive team asks for pass rates instead of accepting demos, questions what each agent may do alone, asks what untrusted content agents read, reads cost-per-task trends, makes autonomy decisions on cited evidence, and communicates capacity decisions honestly. The AI lead or the facilitator records the questions asked in the two reviews following each module and reports the shift. A short written assessment can supplement this, but the questions in the monthly review are the real test.

Who should deliver it?

Someone who has taken agents to production and can use the company's own agents and evidence as material: the internal AI lead, an experienced practitioner partner, or both together. Vendors with products to sell teach products; academics without production experience teach concepts. Neither converts reliably into the decisions executives have to make. The delivery works best alongside a live agent deployment, because every module then has real material.

What does the curriculum produce?

A leadership team that can read evidence builds a program that produces it. Once executives ask for pass rates, teams build evaluation sets. Once executives ask what an agent could do if compromised, engineers narrow permissions. Once executives decide autonomy on cited evidence, project teams stop deciding it by default. Literacy is not a prerequisite for the program; it is part of the program, and it compounds with it. The AI-native leadership playbook whitepaper describes the habits the curriculum is designed to form.

What should executives ask about their own literacy?

• Could I explain to a colleague what makes an agent different from a chatbot, and why it matters?
• When did I last ask for a pass rate?
• For our highest-tier agent, could I say what it could do if compromised?
• Did I make an autonomy decision last quarter, and what evidence did I cite?
• Could I explain our cost per task on one agent, against its baseline?
• Would our employees say I was honest about what changes?

How can FISTA Solutions help?

FISTA Solutions delivers executive and board AI literacy programs through its AI enablement practice, as a practitioner rather than a vendor, using the client's own agents and evidence as material; and because it builds the client's AI agents alongside, the curriculum runs on live deployments from the first module. Its forward deployed engineers supply the production experience the delivery depends on. Since 2017, FISTA has delivered 150+ projects for 50+ companies across 12+ countries.

To run the eight modules for your leadership team alongside a live deployment, talk to FISTA on WhatsApp, or start with questions executives should ask about AI agents, which is the curriculum's assessment in one page.