AI Literacy Is Not a One-Hour Training Course

1. The Same AI Tool Can Carry Very Different Risks

Consider three employees using the same generative AI assistant.

One uses it to improve the wording of an internal email. Another asks it to identify trends in customer sales. A pricing specialist uses it to recommend a selling price for an industrial spare part.

Technically, all three employees are using the same type of tool. Operationally, they are doing completely different things.

An imperfect sentence in an internal email can usually be corrected without serious consequences. An incorrect commercial recommendation, however, could reduce margin, create inconsistent customer prices or damage trust in an important relationship.

This is why AI literacy cannot be identical for every employee.

It must be connected to the decisions people make, the information they use and the consequences of getting the answer wrong.

2. What AI Literacy Looks Like in Pricing

In pricing work, AI can significantly reduce the time required to clean data, classify materials, compare historical sales, identify unusual price movements and organise market information.

However, none of these capabilities removes the need for commercial judgment.

Imagine that an AI system finds a publicly advertised price for what appears to be the same spare part. At first glance, this looks like valuable market evidence. Before using it, the pricing professional must still determine:

• Whether it is genuinely the same manufacturer part number
• Whether the price is for one unit or a complete package
• Whether minimum-order quantities apply
• Whether freight, taxes or additional charges are included
• Whether the item is new, repaired, refurbished or obsolete
• Whether the supplier and market are comparable
• Whether the published price is still current

Someone may understand how to operate an AI tool and still accept the wrong comparison.

A truly AI-literate pricing professional understands both the technology and the business context. That person knows when an output is useful, when it requires further investigation and when it should be rejected.

3. AI Can Process the Data, but It Does Not Own the Decision

Pricing decisions rarely come from a single number.

A recommended selling price may depend on purchase price, internal handling costs, intercompany factors, material classification, freight, sales history, expected volume, customer agreements and market position.

AI can help bring this information together. It can highlight missing fields, identify unusual deviations and prepare possible pricing scenarios. However, the final decision still requires someone to understand how those factors interact.

For example, a price may appear expensive compared with a public online offer but remain commercially justified because the industrial customer is paying for assured availability, technical compatibility and service support.

The opposite is also possible. A price calculated correctly according to internal costing logic may still be commercially unrealistic if it is significantly above credible market alternatives.

AI literacy means understanding this tension.

It is not simply knowing how to calculate a price. It is knowing how to challenge the calculation before the price reaches the customer.

4. The Importance of Reliable Business Data

AI performance depends heavily on the quality of the information it receives.

This becomes clear when working with information from enterprise systems such as SAP. Material descriptions may be incomplete, units of measure may differ and supplier records may not always reflect the latest commercial reality. Historical prices may also contain exceptional project conditions that should not be used as the basis for future decisions.

AI can process this information quickly, but speed does not correct poor data automatically.

If an incorrect purchase price enters the calculation, AI can produce a perfectly structured but commercially incorrect recommendation. If a material is classified incorrectly, the wrong pricing logic may be applied consistently across hundreds of items.

This creates a dangerous impression of accuracy. The output looks professional because the calculation is automated, but its foundation may still be wrong.

AI-literate employees therefore need to understand data lineage:

• Where did the information originate?
• When was it last updated?
• Is it complete?
• Does it refer to the correct material, supplier and plant?
• Were any exceptional business conditions included?
• Which assumptions were made during the calculation?

Without these checks, automation can scale errors just as efficiently as it scales correct decisions.

5. A Practical Three-Level AI Literacy Model

Organisations do not need to turn every employee into an AI expert. They need to provide the right knowledge for the right level of responsibility.

Level 1: General users

Everyone using AI should understand the basic rules:

• Do not enter confidential, commercially sensitive or personal information into unapproved tools.
• Verify important facts instead of assuming confident language is accurate.
• Recognise that AI may invent sources or misunderstand instructions.
• Follow company rules on approved systems and acceptable use.
• Remain responsible for any content submitted under their name.

This is the foundation of AI literacy, but it is not sufficient for employees using AI in operational or commercial decisions.

Level 2: Functional specialists

Pricing, procurement, finance, HR and commercial teams require training connected to their professional responsibilities.

For pricing teams, this should include:

• Validating source data and units of measure
• Checking material numbers and manufacturer references
• Distinguishing genuine market comparisons from misleading matches
• Understanding calculation logic and approval limits
• Identifying customer-specific conditions
• Documenting assumptions and exceptions
• Knowing when human review is mandatory

The objective is not only to prevent obvious hallucinations. It is also to prevent a technically plausible answer from becoming a commercially incorrect decision.

Level 3: Process owners and leaders

Managers and process owners require a broader form of AI literacy.

They need to decide:

• Which processes are suitable for AI assistance
• Where human approval must remain
• Who is accountable when an output is wrong
• How exceptions and corrections will be documented
• Which performance measures demonstrate real value
• How employees will continue developing professional judgment
• Whether faster output is producing better business results

This leadership level is often overlooked. Yet leadership decisions determine whether AI becomes a responsible organisational capability or simply another uncontrolled tool.

6. Human Oversight Must Be Meaningful

Many organisations describe their AI processes as “human-in-the-loop.”

However, adding a human approval button does not automatically create effective oversight.

If an employee receives hundreds of AI-generated recommendations and is expected to approve them quickly, that person may simply accept the output without properly reviewing it. Human oversight then exists on paper but not in practice.

Meaningful oversight requires:

• Enough information to understand the recommendation
• Visibility of the source data and important assumptions
• Clear explanations for unusual results
• Sufficient time to challenge the output
• Authority to reject or change the recommendation
• Documentation of significant exceptions

In a pricing process, the reviewer should be able to see why a price has changed—not only the final number.

Was the increase caused by a higher purchase price? Did the classification change? Was freight added? Did the system identify a market comparison? Was expected volume adjusted?

Without this transparency, the human reviewer is not truly overseeing the system. The person is only confirming its output.

7. Do Not Measure AI Adoption Only by Usage

Companies often measure AI progress through licences, active users, prompts or training completion rates.

These figures show activity, but they do not prove business value.

A pricing function should instead examine outcomes such as:

• Reduction in turnaround time
• Fewer manual data errors
• Improved price consistency
• Better identification of market exceptions
• More time available for strategic analysis
• Stronger documentation of commercial decisions
• Margin or volume improvement following human-approved actions

The most important question should not be:

“How many employees are using AI?”

It should be:

“Which decisions are becoming faster, better or more reliable—and how do we know?”

This shift from usage to outcomes is essential. Otherwise, organisations may celebrate high adoption while employees quietly spend additional time correcting poor-quality AI output.

8. The Risk of Removing the Learning Process

There is another reason why role-specific AI literacy matters.

Experienced professionals can often recognise when an AI-generated answer does not make sense because they have encountered similar situations before. They have worked through incomplete data, incorrect master records, unusual customer agreements and commercial exceptions.

Junior employees have not yet developed the same pattern recognition.

If AI immediately supplies every calculation, explanation and recommendation, employees may complete tasks faster without building the judgment required to validate the result.

This creates a long-term risk. The organisation gains short-term productivity but gradually weakens the development of its future experts.

Leaders therefore need to protect the learning process.

Employees should sometimes be asked to:

• Explain the reasoning behind a recommendation
• Identify the most important risks
• Calculate a representative example manually
• Compare the AI result with historical outcomes
• Describe which assumptions could change the decision
• Present reasons for accepting or rejecting the output

The purpose is not to slow down AI adoption. It is to ensure that productivity today does not create an expertise gap tomorrow.

9. AI Literacy Requires Continuous Learning

A single training session quickly becomes outdated because AI tools, organisational policies and regulatory requirements continue to change.

AI literacy should therefore be treated as an ongoing capability.

A practical programme could include:

• Basic onboarding for all approved AI users
• Role-specific training for functional teams
• Short updates when tools or company policies change
• Examples of errors and lessons learned
• Regular discussions about new use cases
• A channel where employees can raise questions
• Periodic reviews of high-impact AI-supported processes

Real examples are particularly valuable.

Employees learn more from seeing how an incorrect market comparison was identified or how poor source data changed a pricing recommendation than from reading a long list of general principles.

Organisations should create an environment where people can discuss AI errors openly. If every mistake is treated as personal failure, employees may hide problems instead of helping the organisation improve its controls.

10. From Compliance Requirement to Business Capability

The EU AI Act gives companies an additional reason to take AI literacy seriously. But compliance should not be the only motivation.

Well-designed AI literacy can improve decision quality, reduce operational risk and help organisations capture more value from the systems they already have.

The European Commission does not prescribe one identical course for everyone. Its guidance supports a contextual and risk-based approach. This gives organisations an opportunity to move beyond generic training and connect AI literacy directly to real work.

For a pricing team, that means learning how to validate market evidence, protect commercial logic and document human oversight.

For general employees, it means understanding confidentiality, verification and individual responsibility.

For leaders, it means knowing where AI can assist, where it must be challenged and where accountability must remain unmistakably human.

Conclusion

AI literacy is not the ability to produce an answer with a prompt.

It is the ability to understand whether that answer is reliable, appropriate and suitable for influencing a real decision.

A one-hour training session may introduce the technology, but it cannot prepare every employee for every business context. Meaningful AI literacy must be role-specific, risk-based and connected to actual workflows.

The organisations that understand this will not only be better prepared for regulatory expectations. They will also make better decisions, preserve professional judgment and create more sustainable value from AI.

The goal should not be to make every employee an AI expert.

The goal should be to ensure that everyone using AI understands what they are responsible for.

Source and further reading: European Commission—AI Literacy Questions and Answers, updated July 2026.

This article offers a practical business interpretation of AI literacy and does not constitute legal advice.